xref: /qemu/include/block/block.h (revision 187c6147)
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 BdrvChildRole BdrvChildRole;
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     /*
55      * The BDRV_REQ_NO_SERIALISING flag is only valid for reads and means that
56      * we don't want wait_serialising_requests() during the read operation.
57      *
58      * This flag is used for backup copy-on-write operations, when we need to
59      * read old data before write (write notifier triggered). It is okay since
60      * we already waited for other serializing requests in the initiating write
61      * (see bdrv_aligned_pwritev), and it is necessary if the initiating write
62      * is already serializing (without the flag, the read would deadlock
63      * waiting for the serialising write to complete).
64      */
65     BDRV_REQ_NO_SERIALISING     = 0x8,
66     BDRV_REQ_FUA                = 0x10,
67     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
68 
69     /* Signifies that this write request will not change the visible disk
70      * content. */
71     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
72 
73     /*
74      * BDRV_REQ_SERIALISING forces request serialisation for writes.
75      * It is used to ensure that writes to the backing file of a backup process
76      * target cannot race with a read of the backup target that defers to the
77      * backing file.
78      *
79      * Note, that BDRV_REQ_SERIALISING is _not_ opposite in meaning to
80      * BDRV_REQ_NO_SERIALISING. A more descriptive name for the latter might be
81      * _DO_NOT_WAIT_FOR_SERIALISING, except that is too long.
82      */
83     BDRV_REQ_SERIALISING        = 0x80,
84 
85     /* Execute the request only if the operation can be offloaded or otherwise
86      * be executed efficiently, but return an error instead of using a slow
87      * fallback. */
88     BDRV_REQ_NO_FALLBACK        = 0x100,
89 
90     /*
91      * BDRV_REQ_PREFETCH may be used only together with BDRV_REQ_COPY_ON_READ
92      * on read request and means that caller doesn't really need data to be
93      * written to qiov parameter which may be NULL.
94      */
95     BDRV_REQ_PREFETCH  = 0x200,
96     /* Mask of valid flags */
97     BDRV_REQ_MASK               = 0x3ff,
98 } BdrvRequestFlags;
99 
100 typedef struct BlockSizes {
101     uint32_t phys;
102     uint32_t log;
103 } BlockSizes;
104 
105 typedef struct HDGeometry {
106     uint32_t heads;
107     uint32_t sectors;
108     uint32_t cylinders;
109 } HDGeometry;
110 
111 #define BDRV_O_RDWR        0x0002
112 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
113 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
114 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
115 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
116 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
117 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
118 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
119 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
120 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
121 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
122 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
123 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
124 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
125                                       select an appropriate protocol driver,
126                                       ignoring the format layer */
127 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
128 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
129 
130 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
131 
132 
133 /* Option names of options parsed by the block layer */
134 
135 #define BDRV_OPT_CACHE_WB       "cache.writeback"
136 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
137 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
138 #define BDRV_OPT_READ_ONLY      "read-only"
139 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
140 #define BDRV_OPT_DISCARD        "discard"
141 #define BDRV_OPT_FORCE_SHARE    "force-share"
142 
143 
144 #define BDRV_SECTOR_BITS   9
145 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
146 
147 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
148                                      INT_MAX >> BDRV_SECTOR_BITS)
149 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
150 
151 /*
152  * Allocation status flags for bdrv_block_status() and friends.
153  *
154  * Public flags:
155  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
156  * BDRV_BLOCK_ZERO: offset reads as zero
157  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
158  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
159  *                       layer rather than any backing, set by block layer
160  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
161  *                 layer, set by block layer
162  *
163  * Internal flags:
164  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
165  *                 that the block layer recompute the answer from the returned
166  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
167  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
168  *                     zeroes in file child of current block node inside
169  *                     returned region. Only valid together with both
170  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
171  *                     appear with BDRV_BLOCK_ZERO.
172  *
173  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
174  * host offset within the returned BDS that is allocated for the
175  * corresponding raw guest data.  However, whether that offset
176  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
177  *
178  * DATA ZERO OFFSET_VALID
179  *  t    t        t       sectors read as zero, returned file is zero at offset
180  *  t    f        t       sectors read as valid from file at offset
181  *  f    t        t       sectors preallocated, read as zero, returned file not
182  *                        necessarily zero at offset
183  *  f    f        t       sectors preallocated but read from backing_hd,
184  *                        returned file contains garbage at offset
185  *  t    t        f       sectors preallocated, read as zero, unknown offset
186  *  t    f        f       sectors read from unknown file or offset
187  *  f    t        f       not allocated or unknown offset, read as zero
188  *  f    f        f       not allocated or unknown offset, read from backing_hd
189  */
190 #define BDRV_BLOCK_DATA         0x01
191 #define BDRV_BLOCK_ZERO         0x02
192 #define BDRV_BLOCK_OFFSET_VALID 0x04
193 #define BDRV_BLOCK_RAW          0x08
194 #define BDRV_BLOCK_ALLOCATED    0x10
195 #define BDRV_BLOCK_EOF          0x20
196 #define BDRV_BLOCK_RECURSE      0x40
197 
198 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
199 
200 typedef struct BDRVReopenState {
201     BlockDriverState *bs;
202     int flags;
203     BlockdevDetectZeroesOptions detect_zeroes;
204     bool backing_missing;
205     bool replace_backing_bs;  /* new_backing_bs is ignored if this is false */
206     BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */
207     uint64_t perm, shared_perm;
208     QDict *options;
209     QDict *explicit_options;
210     void *opaque;
211 } BDRVReopenState;
212 
213 /*
214  * Block operation types
215  */
216 typedef enum BlockOpType {
217     BLOCK_OP_TYPE_BACKUP_SOURCE,
218     BLOCK_OP_TYPE_BACKUP_TARGET,
219     BLOCK_OP_TYPE_CHANGE,
220     BLOCK_OP_TYPE_COMMIT_SOURCE,
221     BLOCK_OP_TYPE_COMMIT_TARGET,
222     BLOCK_OP_TYPE_DATAPLANE,
223     BLOCK_OP_TYPE_DRIVE_DEL,
224     BLOCK_OP_TYPE_EJECT,
225     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
226     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
227     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
228     BLOCK_OP_TYPE_MIRROR_SOURCE,
229     BLOCK_OP_TYPE_MIRROR_TARGET,
230     BLOCK_OP_TYPE_RESIZE,
231     BLOCK_OP_TYPE_STREAM,
232     BLOCK_OP_TYPE_REPLACE,
233     BLOCK_OP_TYPE_MAX,
234 } BlockOpType;
235 
236 /* Block node permission constants */
237 enum {
238     /**
239      * A user that has the "permission" of consistent reads is guaranteed that
240      * their view of the contents of the block device is complete and
241      * self-consistent, representing the contents of a disk at a specific
242      * point.
243      *
244      * For most block devices (including their backing files) this is true, but
245      * the property cannot be maintained in a few situations like for
246      * intermediate nodes of a commit block job.
247      */
248     BLK_PERM_CONSISTENT_READ    = 0x01,
249 
250     /** This permission is required to change the visible disk contents. */
251     BLK_PERM_WRITE              = 0x02,
252 
253     /**
254      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
255      * required for writes to the block node when the caller promises that
256      * the visible disk content doesn't change.
257      *
258      * As the BLK_PERM_WRITE permission is strictly stronger, either is
259      * sufficient to perform an unchanging write.
260      */
261     BLK_PERM_WRITE_UNCHANGED    = 0x04,
262 
263     /** This permission is required to change the size of a block node. */
264     BLK_PERM_RESIZE             = 0x08,
265 
266     /**
267      * This permission is required to change the node that this BdrvChild
268      * points to.
269      */
270     BLK_PERM_GRAPH_MOD          = 0x10,
271 
272     BLK_PERM_ALL                = 0x1f,
273 
274     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
275                                  | BLK_PERM_WRITE
276                                  | BLK_PERM_WRITE_UNCHANGED
277                                  | BLK_PERM_RESIZE,
278 
279     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
280 };
281 
282 char *bdrv_perm_names(uint64_t perm);
283 
284 /* disk I/O throttling */
285 void bdrv_init(void);
286 void bdrv_init_with_whitelist(void);
287 bool bdrv_uses_whitelist(void);
288 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
289 BlockDriver *bdrv_find_protocol(const char *filename,
290                                 bool allow_protocol_prefix,
291                                 Error **errp);
292 BlockDriver *bdrv_find_format(const char *format_name);
293 int bdrv_create(BlockDriver *drv, const char* filename,
294                 QemuOpts *opts, Error **errp);
295 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
296 BlockDriverState *bdrv_new(void);
297 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
298                  Error **errp);
299 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
300                        Error **errp);
301 
302 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
303 int bdrv_parse_discard_flags(const char *mode, int *flags);
304 BdrvChild *bdrv_open_child(const char *filename,
305                            QDict *options, const char *bdref_key,
306                            BlockDriverState* parent,
307                            const BdrvChildRole *child_role,
308                            bool allow_none, Error **errp);
309 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
310 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
311                          Error **errp);
312 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
313                            const char *bdref_key, Error **errp);
314 BlockDriverState *bdrv_open(const char *filename, const char *reference,
315                             QDict *options, int flags, Error **errp);
316 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
317                                        int flags, Error **errp);
318 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
319                                     BlockDriverState *bs, QDict *options,
320                                     bool keep_old_opts);
321 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
322 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
323                               Error **errp);
324 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
325                         BlockReopenQueue *queue, Error **errp);
326 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
327 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
328 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
329                        int bytes, BdrvRequestFlags flags);
330 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
331 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
332 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
333 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
334 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
335 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
336                      const void *buf, int count);
337 /*
338  * Efficiently zero a region of the disk image.  Note that this is a regular
339  * I/O request like read or write and should have a reasonable size.  This
340  * function is not suitable for zeroing the entire image in a single request
341  * because it may allocate memory for the entire region.
342  */
343 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
344                                        int bytes, BdrvRequestFlags flags);
345 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
346     const char *backing_file);
347 void bdrv_refresh_filename(BlockDriverState *bs);
348 
349 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
350                                   PreallocMode prealloc, Error **errp);
351 int bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
352                   PreallocMode prealloc, Error **errp);
353 
354 int64_t bdrv_nb_sectors(BlockDriverState *bs);
355 int64_t bdrv_getlength(BlockDriverState *bs);
356 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
357 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
358                                BlockDriverState *in_bs, Error **errp);
359 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
360 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
361 int bdrv_commit(BlockDriverState *bs);
362 int bdrv_change_backing_file(BlockDriverState *bs,
363     const char *backing_file, const char *backing_fmt);
364 void bdrv_register(BlockDriver *bdrv);
365 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
366                            const char *backing_file_str);
367 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
368                                     BlockDriverState *bs);
369 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
370 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
371                                   Error **errp);
372 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
373                               Error **errp);
374 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
375 
376 
377 typedef struct BdrvCheckResult {
378     int corruptions;
379     int leaks;
380     int check_errors;
381     int corruptions_fixed;
382     int leaks_fixed;
383     int64_t image_end_offset;
384     BlockFragInfo bfi;
385 } BdrvCheckResult;
386 
387 typedef enum {
388     BDRV_FIX_LEAKS    = 1,
389     BDRV_FIX_ERRORS   = 2,
390 } BdrvCheckMode;
391 
392 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
393 
394 /* The units of offset and total_work_size may be chosen arbitrarily by the
395  * block driver; total_work_size may change during the course of the amendment
396  * operation */
397 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
398                                       int64_t total_work_size, void *opaque);
399 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
400                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
401                        Error **errp);
402 
403 /* external snapshots */
404 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
405                                       BlockDriverState *candidate);
406 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
407 
408 /* check if a named node can be replaced when doing drive-mirror */
409 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
410                                         const char *node_name, Error **errp);
411 
412 /* async block I/O */
413 void bdrv_aio_cancel(BlockAIOCB *acb);
414 void bdrv_aio_cancel_async(BlockAIOCB *acb);
415 
416 /* sg packet commands */
417 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
418 
419 /* Invalidate any cached metadata used by image formats */
420 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
421 void bdrv_invalidate_cache_all(Error **errp);
422 int bdrv_inactivate_all(void);
423 
424 /* Ensure contents are flushed to disk.  */
425 int bdrv_flush(BlockDriverState *bs);
426 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
427 int bdrv_flush_all(void);
428 void bdrv_close_all(void);
429 void bdrv_drain(BlockDriverState *bs);
430 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
431 void bdrv_drain_all_begin(void);
432 void bdrv_drain_all_end(void);
433 void bdrv_drain_all(void);
434 
435 #define BDRV_POLL_WHILE(bs, cond) ({                       \
436     BlockDriverState *bs_ = (bs);                          \
437     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
438                    cond); })
439 
440 int bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
441 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
442 int bdrv_has_zero_init_1(BlockDriverState *bs);
443 int bdrv_has_zero_init(BlockDriverState *bs);
444 int bdrv_has_zero_init_truncate(BlockDriverState *bs);
445 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
446 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
447 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
448                       int64_t bytes, int64_t *pnum, int64_t *map,
449                       BlockDriverState **file);
450 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
451                             int64_t offset, int64_t bytes, int64_t *pnum,
452                             int64_t *map, BlockDriverState **file);
453 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
454                       int64_t *pnum);
455 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
456                             bool include_base, int64_t offset, int64_t bytes,
457                             int64_t *pnum);
458 
459 bool bdrv_is_read_only(BlockDriverState *bs);
460 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
461                            bool ignore_allow_rdw, Error **errp);
462 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
463                               Error **errp);
464 bool bdrv_is_writable(BlockDriverState *bs);
465 bool bdrv_is_sg(BlockDriverState *bs);
466 bool bdrv_is_inserted(BlockDriverState *bs);
467 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
468 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
469 const char *bdrv_get_format_name(BlockDriverState *bs);
470 BlockDriverState *bdrv_find_node(const char *node_name);
471 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
472 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
473 BlockDriverState *bdrv_lookup_bs(const char *device,
474                                  const char *node_name,
475                                  Error **errp);
476 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
477 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
478 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
479 
480 typedef struct BdrvNextIterator {
481     enum {
482         BDRV_NEXT_BACKEND_ROOTS,
483         BDRV_NEXT_MONITOR_OWNED,
484     } phase;
485     BlockBackend *blk;
486     BlockDriverState *bs;
487 } BdrvNextIterator;
488 
489 BlockDriverState *bdrv_first(BdrvNextIterator *it);
490 BlockDriverState *bdrv_next(BdrvNextIterator *it);
491 void bdrv_next_cleanup(BdrvNextIterator *it);
492 
493 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
494 bool bdrv_is_encrypted(BlockDriverState *bs);
495 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
496                          void *opaque, bool read_only);
497 const char *bdrv_get_node_name(const BlockDriverState *bs);
498 const char *bdrv_get_device_name(const BlockDriverState *bs);
499 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
500 int bdrv_get_flags(BlockDriverState *bs);
501 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
502 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
503                                           Error **errp);
504 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
505 void bdrv_round_to_clusters(BlockDriverState *bs,
506                             int64_t offset, int64_t bytes,
507                             int64_t *cluster_offset,
508                             int64_t *cluster_bytes);
509 
510 void bdrv_get_backing_filename(BlockDriverState *bs,
511                                char *filename, int filename_size);
512 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
513 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
514                                                    const char *backing,
515                                                    Error **errp);
516 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
517 
518 int path_has_protocol(const char *path);
519 int path_is_absolute(const char *path);
520 char *path_combine(const char *base_path, const char *filename);
521 
522 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
523 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
524 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
525                       int64_t pos, int size);
526 
527 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
528                       int64_t pos, int size);
529 
530 void bdrv_img_create(const char *filename, const char *fmt,
531                      const char *base_filename, const char *base_fmt,
532                      char *options, uint64_t img_size, int flags,
533                      bool quiet, Error **errp);
534 
535 /* Returns the alignment in bytes that is required so that no bounce buffer
536  * is required throughout the stack */
537 size_t bdrv_min_mem_align(BlockDriverState *bs);
538 /* Returns optimal alignment in bytes for bounce buffer */
539 size_t bdrv_opt_mem_align(BlockDriverState *bs);
540 void *qemu_blockalign(BlockDriverState *bs, size_t size);
541 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
542 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
543 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
544 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
545 
546 void bdrv_enable_copy_on_read(BlockDriverState *bs);
547 void bdrv_disable_copy_on_read(BlockDriverState *bs);
548 
549 void bdrv_ref(BlockDriverState *bs);
550 void bdrv_unref(BlockDriverState *bs);
551 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
552 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
553                              BlockDriverState *child_bs,
554                              const char *child_name,
555                              const BdrvChildRole *child_role,
556                              Error **errp);
557 
558 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
559 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
560 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
561 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
562 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
563 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
564 
565 #define BLKDBG_EVENT(child, evt) \
566     do { \
567         if (child) { \
568             bdrv_debug_event(child->bs, evt); \
569         } \
570     } while (0)
571 
572 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
573 
574 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
575                            const char *tag);
576 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
577 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
578 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
579 
580 /**
581  * bdrv_get_aio_context:
582  *
583  * Returns: the currently bound #AioContext
584  */
585 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
586 
587 /**
588  * Transfer control to @co in the aio context of @bs
589  */
590 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
591 
592 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
593                                  AioContext *new_context, GSList **ignore);
594 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
595                              Error **errp);
596 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
597                                    BdrvChild *ignore_child, Error **errp);
598 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
599                                     GSList **ignore, Error **errp);
600 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
601                               GSList **ignore, Error **errp);
602 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
603 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
604 
605 void bdrv_io_plug(BlockDriverState *bs);
606 void bdrv_io_unplug(BlockDriverState *bs);
607 
608 /**
609  * bdrv_parent_drained_begin_single:
610  *
611  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
612  * any pending activity to cease.
613  */
614 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
615 
616 /**
617  * bdrv_parent_drained_end_single:
618  *
619  * End a quiesced section for the parent of @c.
620  *
621  * This polls @bs's AioContext until all scheduled sub-drained_ends
622  * have settled, which may result in graph changes.
623  */
624 void bdrv_parent_drained_end_single(BdrvChild *c);
625 
626 /**
627  * bdrv_drain_poll:
628  *
629  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
630  * and if @recursive is true its children as well (used for subtree drain).
631  *
632  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
633  * ignore the drain request because they will be drained separately (used for
634  * drain_all).
635  *
636  * This is part of bdrv_drained_begin.
637  */
638 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
639                      BdrvChild *ignore_parent, bool ignore_bds_parents);
640 
641 /**
642  * bdrv_drained_begin:
643  *
644  * Begin a quiesced section for exclusive access to the BDS, by disabling
645  * external request sources including NBD server and device model. Note that
646  * this doesn't block timers or coroutines from submitting more requests, which
647  * means block_job_pause is still necessary.
648  *
649  * This function can be recursive.
650  */
651 void bdrv_drained_begin(BlockDriverState *bs);
652 
653 /**
654  * bdrv_do_drained_begin_quiesce:
655  *
656  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
657  * running requests to complete.
658  */
659 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
660                                    BdrvChild *parent, bool ignore_bds_parents);
661 
662 /**
663  * Like bdrv_drained_begin, but recursively begins a quiesced section for
664  * exclusive access to all child nodes as well.
665  */
666 void bdrv_subtree_drained_begin(BlockDriverState *bs);
667 
668 /**
669  * bdrv_drained_end:
670  *
671  * End a quiescent section started by bdrv_drained_begin().
672  *
673  * This polls @bs's AioContext until all scheduled sub-drained_ends
674  * have settled.  On one hand, that may result in graph changes.  On
675  * the other, this requires that the caller either runs in the main
676  * loop; or that all involved nodes (@bs and all of its parents) are
677  * in the caller's AioContext.
678  */
679 void bdrv_drained_end(BlockDriverState *bs);
680 
681 /**
682  * bdrv_drained_end_no_poll:
683  *
684  * Same as bdrv_drained_end(), but do not poll for the subgraph to
685  * actually become unquiesced.  Therefore, no graph changes will occur
686  * with this function.
687  *
688  * *drained_end_counter is incremented for every background operation
689  * that is scheduled, and will be decremented for every operation once
690  * it settles.  The caller must poll until it reaches 0.  The counter
691  * should be accessed using atomic operations only.
692  */
693 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
694 
695 /**
696  * End a quiescent section started by bdrv_subtree_drained_begin().
697  */
698 void bdrv_subtree_drained_end(BlockDriverState *bs);
699 
700 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
701                     Error **errp);
702 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
703 
704 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
705                                      uint32_t granularity, Error **errp);
706 /**
707  *
708  * bdrv_register_buf/bdrv_unregister_buf:
709  *
710  * Register/unregister a buffer for I/O. For example, VFIO drivers are
711  * interested to know the memory areas that would later be used for I/O, so
712  * that they can prepare IOMMU mapping etc., to get better performance.
713  */
714 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
715 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
716 
717 /**
718  *
719  * bdrv_co_copy_range:
720  *
721  * Do offloaded copy between two children. If the operation is not implemented
722  * by the driver, or if the backend storage doesn't support it, a negative
723  * error code will be returned.
724  *
725  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
726  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
727  * calling copy_file_range(2)) after the first error, thus it should fall back
728  * to a read+write path in the caller level.
729  *
730  * @src: Source child to copy data from
731  * @src_offset: offset in @src image to read data
732  * @dst: Destination child to copy data to
733  * @dst_offset: offset in @dst image to write data
734  * @bytes: number of bytes to copy
735  * @flags: request flags. Supported flags:
736  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
737  *                               write on @dst as if bdrv_co_pwrite_zeroes is
738  *                               called. Used to simplify caller code, or
739  *                               during BlockDriver.bdrv_co_copy_range_from()
740  *                               recursion.
741  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
742  *                                   requests currently in flight.
743  *
744  * Returns: 0 if succeeded; negative error code if failed.
745  **/
746 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
747                                     BdrvChild *dst, uint64_t dst_offset,
748                                     uint64_t bytes, BdrvRequestFlags read_flags,
749                                     BdrvRequestFlags write_flags);
750 #endif
751