xref: /openbmc/qemu/include/block/block.h (revision 0c4e9931)
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 /* block.c */
14 typedef struct BlockDriver BlockDriver;
15 typedef struct BdrvChild BdrvChild;
16 typedef struct BdrvChildClass BdrvChildClass;
17 
18 typedef struct BlockDriverInfo {
19     /* in bytes, 0 if irrelevant */
20     int cluster_size;
21     /* offset at which the VM state can be saved (0 if not possible) */
22     int64_t vm_state_offset;
23     bool is_dirty;
24     /*
25      * True if unallocated blocks read back as zeroes. This is equivalent
26      * to the LBPRZ flag in the SCSI logical block provisioning page.
27      */
28     bool unallocated_blocks_are_zero;
29     /*
30      * True if this block driver only supports compressed writes
31      */
32     bool needs_compressed_writes;
33 } BlockDriverInfo;
34 
35 typedef struct BlockFragInfo {
36     uint64_t allocated_clusters;
37     uint64_t total_clusters;
38     uint64_t fragmented_clusters;
39     uint64_t compressed_clusters;
40 } BlockFragInfo;
41 
42 typedef enum {
43     BDRV_REQ_COPY_ON_READ       = 0x1,
44     BDRV_REQ_ZERO_WRITE         = 0x2,
45 
46     /*
47      * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
48      * that the block driver should unmap (discard) blocks if it is guaranteed
49      * that the result will read back as zeroes. The flag is only passed to the
50      * driver if the block device is opened with BDRV_O_UNMAP.
51      */
52     BDRV_REQ_MAY_UNMAP          = 0x4,
53 
54     BDRV_REQ_FUA                = 0x10,
55     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
56 
57     /* Signifies that this write request will not change the visible disk
58      * content. */
59     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
60 
61     /*
62      * BDRV_REQ_SERIALISING forces request serialisation for writes.
63      * It is used to ensure that writes to the backing file of a backup process
64      * target cannot race with a read of the backup target that defers to the
65      * backing file.
66      *
67      * Note, that BDRV_REQ_SERIALISING is _not_ opposite in meaning to
68      * BDRV_REQ_NO_SERIALISING. A more descriptive name for the latter might be
69      * _DO_NOT_WAIT_FOR_SERIALISING, except that is too long.
70      */
71     BDRV_REQ_SERIALISING        = 0x80,
72 
73     /* Execute the request only if the operation can be offloaded or otherwise
74      * be executed efficiently, but return an error instead of using a slow
75      * fallback. */
76     BDRV_REQ_NO_FALLBACK        = 0x100,
77 
78     /*
79      * BDRV_REQ_PREFETCH may be used only together with BDRV_REQ_COPY_ON_READ
80      * on read request and means that caller doesn't really need data to be
81      * written to qiov parameter which may be NULL.
82      */
83     BDRV_REQ_PREFETCH  = 0x200,
84     /* Mask of valid flags */
85     BDRV_REQ_MASK               = 0x3ff,
86 } BdrvRequestFlags;
87 
88 typedef struct BlockSizes {
89     uint32_t phys;
90     uint32_t log;
91 } BlockSizes;
92 
93 typedef struct HDGeometry {
94     uint32_t heads;
95     uint32_t sectors;
96     uint32_t cylinders;
97 } HDGeometry;
98 
99 #define BDRV_O_RDWR        0x0002
100 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
101 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
102 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
103 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
104 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
105 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
106 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
107 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
108 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
109 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
110 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
111 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
112 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
113                                       select an appropriate protocol driver,
114                                       ignoring the format layer */
115 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
116 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
117 #define BDRV_O_IO_URING    0x40000 /* use io_uring instead of the thread pool */
118 
119 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
120 
121 
122 /* Option names of options parsed by the block layer */
123 
124 #define BDRV_OPT_CACHE_WB       "cache.writeback"
125 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
126 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
127 #define BDRV_OPT_READ_ONLY      "read-only"
128 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
129 #define BDRV_OPT_DISCARD        "discard"
130 #define BDRV_OPT_FORCE_SHARE    "force-share"
131 
132 
133 #define BDRV_SECTOR_BITS   9
134 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
135 
136 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
137                                      INT_MAX >> BDRV_SECTOR_BITS)
138 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
139 
140 /*
141  * Allocation status flags for bdrv_block_status() and friends.
142  *
143  * Public flags:
144  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
145  * BDRV_BLOCK_ZERO: offset reads as zero
146  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
147  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
148  *                       layer rather than any backing, set by block layer
149  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
150  *                 layer, set by block layer
151  *
152  * Internal flags:
153  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
154  *                 that the block layer recompute the answer from the returned
155  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
156  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
157  *                     zeroes in file child of current block node inside
158  *                     returned region. Only valid together with both
159  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
160  *                     appear with BDRV_BLOCK_ZERO.
161  *
162  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
163  * host offset within the returned BDS that is allocated for the
164  * corresponding raw guest data.  However, whether that offset
165  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
166  *
167  * DATA ZERO OFFSET_VALID
168  *  t    t        t       sectors read as zero, returned file is zero at offset
169  *  t    f        t       sectors read as valid from file at offset
170  *  f    t        t       sectors preallocated, read as zero, returned file not
171  *                        necessarily zero at offset
172  *  f    f        t       sectors preallocated but read from backing_hd,
173  *                        returned file contains garbage at offset
174  *  t    t        f       sectors preallocated, read as zero, unknown offset
175  *  t    f        f       sectors read from unknown file or offset
176  *  f    t        f       not allocated or unknown offset, read as zero
177  *  f    f        f       not allocated or unknown offset, read from backing_hd
178  */
179 #define BDRV_BLOCK_DATA         0x01
180 #define BDRV_BLOCK_ZERO         0x02
181 #define BDRV_BLOCK_OFFSET_VALID 0x04
182 #define BDRV_BLOCK_RAW          0x08
183 #define BDRV_BLOCK_ALLOCATED    0x10
184 #define BDRV_BLOCK_EOF          0x20
185 #define BDRV_BLOCK_RECURSE      0x40
186 
187 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
188 
189 typedef struct BDRVReopenState {
190     BlockDriverState *bs;
191     int flags;
192     BlockdevDetectZeroesOptions detect_zeroes;
193     bool backing_missing;
194     bool replace_backing_bs;  /* new_backing_bs is ignored if this is false */
195     BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */
196     uint64_t perm, shared_perm;
197     QDict *options;
198     QDict *explicit_options;
199     void *opaque;
200 } BDRVReopenState;
201 
202 /*
203  * Block operation types
204  */
205 typedef enum BlockOpType {
206     BLOCK_OP_TYPE_BACKUP_SOURCE,
207     BLOCK_OP_TYPE_BACKUP_TARGET,
208     BLOCK_OP_TYPE_CHANGE,
209     BLOCK_OP_TYPE_COMMIT_SOURCE,
210     BLOCK_OP_TYPE_COMMIT_TARGET,
211     BLOCK_OP_TYPE_DATAPLANE,
212     BLOCK_OP_TYPE_DRIVE_DEL,
213     BLOCK_OP_TYPE_EJECT,
214     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
215     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
216     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
217     BLOCK_OP_TYPE_MIRROR_SOURCE,
218     BLOCK_OP_TYPE_MIRROR_TARGET,
219     BLOCK_OP_TYPE_RESIZE,
220     BLOCK_OP_TYPE_STREAM,
221     BLOCK_OP_TYPE_REPLACE,
222     BLOCK_OP_TYPE_MAX,
223 } BlockOpType;
224 
225 /* Block node permission constants */
226 enum {
227     /**
228      * A user that has the "permission" of consistent reads is guaranteed that
229      * their view of the contents of the block device is complete and
230      * self-consistent, representing the contents of a disk at a specific
231      * point.
232      *
233      * For most block devices (including their backing files) this is true, but
234      * the property cannot be maintained in a few situations like for
235      * intermediate nodes of a commit block job.
236      */
237     BLK_PERM_CONSISTENT_READ    = 0x01,
238 
239     /** This permission is required to change the visible disk contents. */
240     BLK_PERM_WRITE              = 0x02,
241 
242     /**
243      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
244      * required for writes to the block node when the caller promises that
245      * the visible disk content doesn't change.
246      *
247      * As the BLK_PERM_WRITE permission is strictly stronger, either is
248      * sufficient to perform an unchanging write.
249      */
250     BLK_PERM_WRITE_UNCHANGED    = 0x04,
251 
252     /** This permission is required to change the size of a block node. */
253     BLK_PERM_RESIZE             = 0x08,
254 
255     /**
256      * This permission is required to change the node that this BdrvChild
257      * points to.
258      */
259     BLK_PERM_GRAPH_MOD          = 0x10,
260 
261     BLK_PERM_ALL                = 0x1f,
262 
263     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
264                                  | BLK_PERM_WRITE
265                                  | BLK_PERM_WRITE_UNCHANGED
266                                  | BLK_PERM_RESIZE,
267 
268     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
269 };
270 
271 /*
272  * Flags that parent nodes assign to child nodes to specify what kind of
273  * role(s) they take.
274  *
275  * At least one of DATA, METADATA, FILTERED, or COW must be set for
276  * every child.
277  */
278 enum BdrvChildRoleBits {
279     /*
280      * This child stores data.
281      * Any node may have an arbitrary number of such children.
282      */
283     BDRV_CHILD_DATA         = (1 << 0),
284 
285     /*
286      * This child stores metadata.
287      * Any node may have an arbitrary number of metadata-storing
288      * children.
289      */
290     BDRV_CHILD_METADATA     = (1 << 1),
291 
292     /*
293      * A child that always presents exactly the same visible data as
294      * the parent, e.g. by virtue of the parent forwarding all reads
295      * and writes.
296      * This flag is mutually exclusive with DATA, METADATA, and COW.
297      * Any node may have at most one filtered child at a time.
298      */
299     BDRV_CHILD_FILTERED     = (1 << 2),
300 
301     /*
302      * Child from which to read all data that isn’t allocated in the
303      * parent (i.e., the backing child); such data is copied to the
304      * parent through COW (and optionally COR).
305      * This field is mutually exclusive with DATA, METADATA, and
306      * FILTERED.
307      * Any node may have at most one such backing child at a time.
308      */
309     BDRV_CHILD_COW          = (1 << 3),
310 
311     /*
312      * The primary child.  For most drivers, this is the child whose
313      * filename applies best to the parent node.
314      * Any node may have at most one primary child at a time.
315      */
316     BDRV_CHILD_PRIMARY      = (1 << 4),
317 
318     /* Useful combination of flags */
319     BDRV_CHILD_IMAGE        = BDRV_CHILD_DATA
320                               | BDRV_CHILD_METADATA
321                               | BDRV_CHILD_PRIMARY,
322 };
323 
324 /* Mask of BdrvChildRoleBits values */
325 typedef unsigned int BdrvChildRole;
326 
327 char *bdrv_perm_names(uint64_t perm);
328 uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm);
329 
330 /* disk I/O throttling */
331 void bdrv_init(void);
332 void bdrv_init_with_whitelist(void);
333 bool bdrv_uses_whitelist(void);
334 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
335 BlockDriver *bdrv_find_protocol(const char *filename,
336                                 bool allow_protocol_prefix,
337                                 Error **errp);
338 BlockDriver *bdrv_find_format(const char *format_name);
339 int bdrv_create(BlockDriver *drv, const char* filename,
340                 QemuOpts *opts, Error **errp);
341 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
342 
343 BlockDriverState *bdrv_new(void);
344 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
345                  Error **errp);
346 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
347                        Error **errp);
348 
349 int bdrv_parse_aio(const char *mode, int *flags);
350 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
351 int bdrv_parse_discard_flags(const char *mode, int *flags);
352 BdrvChild *bdrv_open_child(const char *filename,
353                            QDict *options, const char *bdref_key,
354                            BlockDriverState* parent,
355                            const BdrvChildClass *child_class,
356                            BdrvChildRole child_role,
357                            bool allow_none, Error **errp);
358 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
359 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
360                          Error **errp);
361 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
362                            const char *bdref_key, Error **errp);
363 BlockDriverState *bdrv_open(const char *filename, const char *reference,
364                             QDict *options, int flags, Error **errp);
365 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
366                                        int flags, Error **errp);
367 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
368                                     BlockDriverState *bs, QDict *options,
369                                     bool keep_old_opts);
370 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
371 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
372                               Error **errp);
373 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
374                         BlockReopenQueue *queue, Error **errp);
375 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
376 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
377 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
378                        int bytes, BdrvRequestFlags flags);
379 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
380 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
381 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
382 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
383 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
384 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
385                      const void *buf, int count);
386 /*
387  * Efficiently zero a region of the disk image.  Note that this is a regular
388  * I/O request like read or write and should have a reasonable size.  This
389  * function is not suitable for zeroing the entire image in a single request
390  * because it may allocate memory for the entire region.
391  */
392 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
393                                        int bytes, BdrvRequestFlags flags);
394 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
395     const char *backing_file);
396 void bdrv_refresh_filename(BlockDriverState *bs);
397 
398 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
399                                   PreallocMode prealloc, BdrvRequestFlags flags,
400                                   Error **errp);
401 int bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
402                   PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
403 
404 int64_t bdrv_nb_sectors(BlockDriverState *bs);
405 int64_t bdrv_getlength(BlockDriverState *bs);
406 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
407 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
408                                BlockDriverState *in_bs, Error **errp);
409 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
410 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
411 int bdrv_commit(BlockDriverState *bs);
412 int bdrv_make_empty(BdrvChild *c, Error **errp);
413 int bdrv_change_backing_file(BlockDriverState *bs,
414     const char *backing_file, const char *backing_fmt);
415 void bdrv_register(BlockDriver *bdrv);
416 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
417                            const char *backing_file_str);
418 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
419                                     BlockDriverState *bs);
420 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
421 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
422                                   Error **errp);
423 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
424                               Error **errp);
425 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
426 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
427 
428 
429 typedef struct BdrvCheckResult {
430     int corruptions;
431     int leaks;
432     int check_errors;
433     int corruptions_fixed;
434     int leaks_fixed;
435     int64_t image_end_offset;
436     BlockFragInfo bfi;
437 } BdrvCheckResult;
438 
439 typedef enum {
440     BDRV_FIX_LEAKS    = 1,
441     BDRV_FIX_ERRORS   = 2,
442 } BdrvCheckMode;
443 
444 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
445 
446 /* The units of offset and total_work_size may be chosen arbitrarily by the
447  * block driver; total_work_size may change during the course of the amendment
448  * operation */
449 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
450                                       int64_t total_work_size, void *opaque);
451 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
452                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
453                        Error **errp);
454 
455 /* check if a named node can be replaced when doing drive-mirror */
456 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
457                                         const char *node_name, Error **errp);
458 
459 /* async block I/O */
460 void bdrv_aio_cancel(BlockAIOCB *acb);
461 void bdrv_aio_cancel_async(BlockAIOCB *acb);
462 
463 /* sg packet commands */
464 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
465 
466 /* Invalidate any cached metadata used by image formats */
467 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
468 void bdrv_invalidate_cache_all(Error **errp);
469 int bdrv_inactivate_all(void);
470 
471 /* Ensure contents are flushed to disk.  */
472 int bdrv_flush(BlockDriverState *bs);
473 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
474 int bdrv_flush_all(void);
475 void bdrv_close_all(void);
476 void bdrv_drain(BlockDriverState *bs);
477 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
478 void bdrv_drain_all_begin(void);
479 void bdrv_drain_all_end(void);
480 void bdrv_drain_all(void);
481 
482 #define BDRV_POLL_WHILE(bs, cond) ({                       \
483     BlockDriverState *bs_ = (bs);                          \
484     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
485                    cond); })
486 
487 int bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
488 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
489 int bdrv_has_zero_init_1(BlockDriverState *bs);
490 int bdrv_has_zero_init(BlockDriverState *bs);
491 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
492 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
493 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
494                       int64_t bytes, int64_t *pnum, int64_t *map,
495                       BlockDriverState **file);
496 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
497                             int64_t offset, int64_t bytes, int64_t *pnum,
498                             int64_t *map, BlockDriverState **file);
499 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
500                       int64_t *pnum);
501 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
502                             bool include_base, int64_t offset, int64_t bytes,
503                             int64_t *pnum);
504 
505 bool bdrv_is_read_only(BlockDriverState *bs);
506 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
507                            bool ignore_allow_rdw, Error **errp);
508 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
509                               Error **errp);
510 bool bdrv_is_writable(BlockDriverState *bs);
511 bool bdrv_is_sg(BlockDriverState *bs);
512 bool bdrv_is_inserted(BlockDriverState *bs);
513 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
514 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
515 const char *bdrv_get_format_name(BlockDriverState *bs);
516 BlockDriverState *bdrv_find_node(const char *node_name);
517 BlockDeviceInfoList *bdrv_named_nodes_list(bool flat, Error **errp);
518 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
519 BlockDriverState *bdrv_lookup_bs(const char *device,
520                                  const char *node_name,
521                                  Error **errp);
522 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
523 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
524 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
525 
526 typedef struct BdrvNextIterator {
527     enum {
528         BDRV_NEXT_BACKEND_ROOTS,
529         BDRV_NEXT_MONITOR_OWNED,
530     } phase;
531     BlockBackend *blk;
532     BlockDriverState *bs;
533 } BdrvNextIterator;
534 
535 BlockDriverState *bdrv_first(BdrvNextIterator *it);
536 BlockDriverState *bdrv_next(BdrvNextIterator *it);
537 void bdrv_next_cleanup(BdrvNextIterator *it);
538 
539 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
540 bool bdrv_is_encrypted(BlockDriverState *bs);
541 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
542                          void *opaque, bool read_only);
543 const char *bdrv_get_node_name(const BlockDriverState *bs);
544 const char *bdrv_get_device_name(const BlockDriverState *bs);
545 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
546 int bdrv_get_flags(BlockDriverState *bs);
547 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
548 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
549                                           Error **errp);
550 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
551 void bdrv_round_to_clusters(BlockDriverState *bs,
552                             int64_t offset, int64_t bytes,
553                             int64_t *cluster_offset,
554                             int64_t *cluster_bytes);
555 
556 void bdrv_get_backing_filename(BlockDriverState *bs,
557                                char *filename, int filename_size);
558 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
559 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
560                                                    const char *backing,
561                                                    Error **errp);
562 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
563 
564 int path_has_protocol(const char *path);
565 int path_is_absolute(const char *path);
566 char *path_combine(const char *base_path, const char *filename);
567 
568 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
569 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
570 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
571                       int64_t pos, int size);
572 
573 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
574                       int64_t pos, int size);
575 
576 void bdrv_img_create(const char *filename, const char *fmt,
577                      const char *base_filename, const char *base_fmt,
578                      char *options, uint64_t img_size, int flags,
579                      bool quiet, Error **errp);
580 
581 /* Returns the alignment in bytes that is required so that no bounce buffer
582  * is required throughout the stack */
583 size_t bdrv_min_mem_align(BlockDriverState *bs);
584 /* Returns optimal alignment in bytes for bounce buffer */
585 size_t bdrv_opt_mem_align(BlockDriverState *bs);
586 void *qemu_blockalign(BlockDriverState *bs, size_t size);
587 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
588 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
589 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
590 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
591 
592 void bdrv_enable_copy_on_read(BlockDriverState *bs);
593 void bdrv_disable_copy_on_read(BlockDriverState *bs);
594 
595 void bdrv_ref(BlockDriverState *bs);
596 void bdrv_unref(BlockDriverState *bs);
597 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
598 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
599                              BlockDriverState *child_bs,
600                              const char *child_name,
601                              const BdrvChildClass *child_class,
602                              BdrvChildRole child_role,
603                              Error **errp);
604 
605 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
606 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
607 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
608 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
609 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
610 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
611 
612 #define BLKDBG_EVENT(child, evt) \
613     do { \
614         if (child) { \
615             bdrv_debug_event(child->bs, evt); \
616         } \
617     } while (0)
618 
619 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
620 
621 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
622                            const char *tag);
623 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
624 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
625 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
626 
627 /**
628  * bdrv_get_aio_context:
629  *
630  * Returns: the currently bound #AioContext
631  */
632 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
633 
634 /**
635  * Transfer control to @co in the aio context of @bs
636  */
637 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
638 
639 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
640                                  AioContext *new_context, GSList **ignore);
641 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
642                              Error **errp);
643 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
644                                    BdrvChild *ignore_child, Error **errp);
645 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
646                                     GSList **ignore, Error **errp);
647 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
648                               GSList **ignore, Error **errp);
649 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
650 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
651 
652 void bdrv_io_plug(BlockDriverState *bs);
653 void bdrv_io_unplug(BlockDriverState *bs);
654 
655 /**
656  * bdrv_parent_drained_begin_single:
657  *
658  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
659  * any pending activity to cease.
660  */
661 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
662 
663 /**
664  * bdrv_parent_drained_end_single:
665  *
666  * End a quiesced section for the parent of @c.
667  *
668  * This polls @bs's AioContext until all scheduled sub-drained_ends
669  * have settled, which may result in graph changes.
670  */
671 void bdrv_parent_drained_end_single(BdrvChild *c);
672 
673 /**
674  * bdrv_drain_poll:
675  *
676  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
677  * and if @recursive is true its children as well (used for subtree drain).
678  *
679  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
680  * ignore the drain request because they will be drained separately (used for
681  * drain_all).
682  *
683  * This is part of bdrv_drained_begin.
684  */
685 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
686                      BdrvChild *ignore_parent, bool ignore_bds_parents);
687 
688 /**
689  * bdrv_drained_begin:
690  *
691  * Begin a quiesced section for exclusive access to the BDS, by disabling
692  * external request sources including NBD server and device model. Note that
693  * this doesn't block timers or coroutines from submitting more requests, which
694  * means block_job_pause is still necessary.
695  *
696  * This function can be recursive.
697  */
698 void bdrv_drained_begin(BlockDriverState *bs);
699 
700 /**
701  * bdrv_do_drained_begin_quiesce:
702  *
703  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
704  * running requests to complete.
705  */
706 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
707                                    BdrvChild *parent, bool ignore_bds_parents);
708 
709 /**
710  * Like bdrv_drained_begin, but recursively begins a quiesced section for
711  * exclusive access to all child nodes as well.
712  */
713 void bdrv_subtree_drained_begin(BlockDriverState *bs);
714 
715 /**
716  * bdrv_drained_end:
717  *
718  * End a quiescent section started by bdrv_drained_begin().
719  *
720  * This polls @bs's AioContext until all scheduled sub-drained_ends
721  * have settled.  On one hand, that may result in graph changes.  On
722  * the other, this requires that the caller either runs in the main
723  * loop; or that all involved nodes (@bs and all of its parents) are
724  * in the caller's AioContext.
725  */
726 void bdrv_drained_end(BlockDriverState *bs);
727 
728 /**
729  * bdrv_drained_end_no_poll:
730  *
731  * Same as bdrv_drained_end(), but do not poll for the subgraph to
732  * actually become unquiesced.  Therefore, no graph changes will occur
733  * with this function.
734  *
735  * *drained_end_counter is incremented for every background operation
736  * that is scheduled, and will be decremented for every operation once
737  * it settles.  The caller must poll until it reaches 0.  The counter
738  * should be accessed using atomic operations only.
739  */
740 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
741 
742 /**
743  * End a quiescent section started by bdrv_subtree_drained_begin().
744  */
745 void bdrv_subtree_drained_end(BlockDriverState *bs);
746 
747 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
748                     Error **errp);
749 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
750 
751 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
752                                      uint32_t granularity, Error **errp);
753 /**
754  *
755  * bdrv_register_buf/bdrv_unregister_buf:
756  *
757  * Register/unregister a buffer for I/O. For example, VFIO drivers are
758  * interested to know the memory areas that would later be used for I/O, so
759  * that they can prepare IOMMU mapping etc., to get better performance.
760  */
761 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
762 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
763 
764 /**
765  *
766  * bdrv_co_copy_range:
767  *
768  * Do offloaded copy between two children. If the operation is not implemented
769  * by the driver, or if the backend storage doesn't support it, a negative
770  * error code will be returned.
771  *
772  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
773  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
774  * calling copy_file_range(2)) after the first error, thus it should fall back
775  * to a read+write path in the caller level.
776  *
777  * @src: Source child to copy data from
778  * @src_offset: offset in @src image to read data
779  * @dst: Destination child to copy data to
780  * @dst_offset: offset in @dst image to write data
781  * @bytes: number of bytes to copy
782  * @flags: request flags. Supported flags:
783  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
784  *                               write on @dst as if bdrv_co_pwrite_zeroes is
785  *                               called. Used to simplify caller code, or
786  *                               during BlockDriver.bdrv_co_copy_range_from()
787  *                               recursion.
788  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
789  *                                   requests currently in flight.
790  *
791  * Returns: 0 if succeeded; negative error code if failed.
792  **/
793 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
794                                     BdrvChild *dst, uint64_t dst_offset,
795                                     uint64_t bytes, BdrvRequestFlags read_flags,
796                                     BdrvRequestFlags write_flags);
797 #endif
798