xref: /openbmc/qemu/include/block/block.h (revision 6a0acfff)
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     /* Mask of valid flags */
91     BDRV_REQ_MASK               = 0x1ff,
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_RDWR        0x0002
106 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
107 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
108 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
109 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
110 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
111 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
112 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
113 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
114 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
115 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
116 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
117 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
118 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
119                                       select an appropriate protocol driver,
120                                       ignoring the format layer */
121 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
122 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
123 
124 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
125 
126 
127 /* Option names of options parsed by the block layer */
128 
129 #define BDRV_OPT_CACHE_WB       "cache.writeback"
130 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
131 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
132 #define BDRV_OPT_READ_ONLY      "read-only"
133 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
134 #define BDRV_OPT_DISCARD        "discard"
135 #define BDRV_OPT_FORCE_SHARE    "force-share"
136 
137 
138 #define BDRV_SECTOR_BITS   9
139 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
140 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
141 
142 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
143                                      INT_MAX >> BDRV_SECTOR_BITS)
144 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
145 
146 /*
147  * Allocation status flags for bdrv_block_status() and friends.
148  *
149  * Public flags:
150  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
151  * BDRV_BLOCK_ZERO: offset reads as zero
152  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
153  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
154  *                       layer rather than any backing, set by block layer
155  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
156  *                 layer, set by block layer
157  *
158  * Internal flags:
159  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
160  *                 that the block layer recompute the answer from the returned
161  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
162  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
163  *                     zeroes in file child of current block node inside
164  *                     returned region. Only valid together with both
165  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
166  *                     appear with BDRV_BLOCK_ZERO.
167  *
168  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
169  * host offset within the returned BDS that is allocated for the
170  * corresponding raw guest data.  However, whether that offset
171  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
172  *
173  * DATA ZERO OFFSET_VALID
174  *  t    t        t       sectors read as zero, returned file is zero at offset
175  *  t    f        t       sectors read as valid from file at offset
176  *  f    t        t       sectors preallocated, read as zero, returned file not
177  *                        necessarily zero at offset
178  *  f    f        t       sectors preallocated but read from backing_hd,
179  *                        returned file contains garbage at offset
180  *  t    t        f       sectors preallocated, read as zero, unknown offset
181  *  t    f        f       sectors read from unknown file or offset
182  *  f    t        f       not allocated or unknown offset, read as zero
183  *  f    f        f       not allocated or unknown offset, read from backing_hd
184  */
185 #define BDRV_BLOCK_DATA         0x01
186 #define BDRV_BLOCK_ZERO         0x02
187 #define BDRV_BLOCK_OFFSET_VALID 0x04
188 #define BDRV_BLOCK_RAW          0x08
189 #define BDRV_BLOCK_ALLOCATED    0x10
190 #define BDRV_BLOCK_EOF          0x20
191 #define BDRV_BLOCK_RECURSE      0x40
192 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
193 
194 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
195 
196 typedef struct BDRVReopenState {
197     BlockDriverState *bs;
198     int flags;
199     BlockdevDetectZeroesOptions detect_zeroes;
200     bool backing_missing;
201     bool replace_backing_bs;  /* new_backing_bs is ignored if this is false */
202     BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */
203     uint64_t perm, shared_perm;
204     QDict *options;
205     QDict *explicit_options;
206     void *opaque;
207 } BDRVReopenState;
208 
209 /*
210  * Block operation types
211  */
212 typedef enum BlockOpType {
213     BLOCK_OP_TYPE_BACKUP_SOURCE,
214     BLOCK_OP_TYPE_BACKUP_TARGET,
215     BLOCK_OP_TYPE_CHANGE,
216     BLOCK_OP_TYPE_COMMIT_SOURCE,
217     BLOCK_OP_TYPE_COMMIT_TARGET,
218     BLOCK_OP_TYPE_DATAPLANE,
219     BLOCK_OP_TYPE_DRIVE_DEL,
220     BLOCK_OP_TYPE_EJECT,
221     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
222     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
223     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
224     BLOCK_OP_TYPE_MIRROR_SOURCE,
225     BLOCK_OP_TYPE_MIRROR_TARGET,
226     BLOCK_OP_TYPE_RESIZE,
227     BLOCK_OP_TYPE_STREAM,
228     BLOCK_OP_TYPE_REPLACE,
229     BLOCK_OP_TYPE_MAX,
230 } BlockOpType;
231 
232 /* Block node permission constants */
233 enum {
234     /**
235      * A user that has the "permission" of consistent reads is guaranteed that
236      * their view of the contents of the block device is complete and
237      * self-consistent, representing the contents of a disk at a specific
238      * point.
239      *
240      * For most block devices (including their backing files) this is true, but
241      * the property cannot be maintained in a few situations like for
242      * intermediate nodes of a commit block job.
243      */
244     BLK_PERM_CONSISTENT_READ    = 0x01,
245 
246     /** This permission is required to change the visible disk contents. */
247     BLK_PERM_WRITE              = 0x02,
248 
249     /**
250      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
251      * required for writes to the block node when the caller promises that
252      * the visible disk content doesn't change.
253      *
254      * As the BLK_PERM_WRITE permission is strictly stronger, either is
255      * sufficient to perform an unchanging write.
256      */
257     BLK_PERM_WRITE_UNCHANGED    = 0x04,
258 
259     /** This permission is required to change the size of a block node. */
260     BLK_PERM_RESIZE             = 0x08,
261 
262     /**
263      * This permission is required to change the node that this BdrvChild
264      * points to.
265      */
266     BLK_PERM_GRAPH_MOD          = 0x10,
267 
268     BLK_PERM_ALL                = 0x1f,
269 
270     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
271                                  | BLK_PERM_WRITE
272                                  | BLK_PERM_WRITE_UNCHANGED
273                                  | BLK_PERM_RESIZE,
274 
275     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
276 };
277 
278 char *bdrv_perm_names(uint64_t perm);
279 
280 /* disk I/O throttling */
281 void bdrv_init(void);
282 void bdrv_init_with_whitelist(void);
283 bool bdrv_uses_whitelist(void);
284 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
285 BlockDriver *bdrv_find_protocol(const char *filename,
286                                 bool allow_protocol_prefix,
287                                 Error **errp);
288 BlockDriver *bdrv_find_format(const char *format_name);
289 int bdrv_create(BlockDriver *drv, const char* filename,
290                 QemuOpts *opts, Error **errp);
291 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
292 BlockDriverState *bdrv_new(void);
293 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
294                  Error **errp);
295 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
296                        Error **errp);
297 
298 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
299 int bdrv_parse_discard_flags(const char *mode, int *flags);
300 BdrvChild *bdrv_open_child(const char *filename,
301                            QDict *options, const char *bdref_key,
302                            BlockDriverState* parent,
303                            const BdrvChildRole *child_role,
304                            bool allow_none, Error **errp);
305 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
306 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
307                          Error **errp);
308 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
309                            const char *bdref_key, Error **errp);
310 BlockDriverState *bdrv_open(const char *filename, const char *reference,
311                             QDict *options, int flags, Error **errp);
312 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
313                                        int flags, Error **errp);
314 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
315                                     BlockDriverState *bs, QDict *options,
316                                     bool keep_old_opts);
317 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
318 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
319                               Error **errp);
320 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
321                         BlockReopenQueue *queue, Error **errp);
322 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
323 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
324 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
325                        int bytes, BdrvRequestFlags flags);
326 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
327 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
328 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
329 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
330 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
331 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
332                      const void *buf, int count);
333 /*
334  * Efficiently zero a region of the disk image.  Note that this is a regular
335  * I/O request like read or write and should have a reasonable size.  This
336  * function is not suitable for zeroing the entire image in a single request
337  * because it may allocate memory for the entire region.
338  */
339 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
340                                        int bytes, BdrvRequestFlags flags);
341 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
342     const char *backing_file);
343 void bdrv_refresh_filename(BlockDriverState *bs);
344 
345 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset,
346                                   PreallocMode prealloc, Error **errp);
347 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
348                   Error **errp);
349 
350 int64_t bdrv_nb_sectors(BlockDriverState *bs);
351 int64_t bdrv_getlength(BlockDriverState *bs);
352 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
353 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
354                                BlockDriverState *in_bs, Error **errp);
355 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
356 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
357 int bdrv_commit(BlockDriverState *bs);
358 int bdrv_change_backing_file(BlockDriverState *bs,
359     const char *backing_file, const char *backing_fmt);
360 void bdrv_register(BlockDriver *bdrv);
361 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
362                            const char *backing_file_str);
363 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
364                                     BlockDriverState *bs);
365 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
366 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
367                                   Error **errp);
368 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
369                               Error **errp);
370 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
371 
372 
373 typedef struct BdrvCheckResult {
374     int corruptions;
375     int leaks;
376     int check_errors;
377     int corruptions_fixed;
378     int leaks_fixed;
379     int64_t image_end_offset;
380     BlockFragInfo bfi;
381 } BdrvCheckResult;
382 
383 typedef enum {
384     BDRV_FIX_LEAKS    = 1,
385     BDRV_FIX_ERRORS   = 2,
386 } BdrvCheckMode;
387 
388 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
389 
390 /* The units of offset and total_work_size may be chosen arbitrarily by the
391  * block driver; total_work_size may change during the course of the amendment
392  * operation */
393 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
394                                       int64_t total_work_size, void *opaque);
395 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
396                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
397                        Error **errp);
398 
399 /* external snapshots */
400 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
401                                       BlockDriverState *candidate);
402 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
403 
404 /* check if a named node can be replaced when doing drive-mirror */
405 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
406                                         const char *node_name, Error **errp);
407 
408 /* async block I/O */
409 void bdrv_aio_cancel(BlockAIOCB *acb);
410 void bdrv_aio_cancel_async(BlockAIOCB *acb);
411 
412 /* sg packet commands */
413 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
414 
415 /* Invalidate any cached metadata used by image formats */
416 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
417 void bdrv_invalidate_cache_all(Error **errp);
418 int bdrv_inactivate_all(void);
419 
420 /* Ensure contents are flushed to disk.  */
421 int bdrv_flush(BlockDriverState *bs);
422 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
423 int bdrv_flush_all(void);
424 void bdrv_close_all(void);
425 void bdrv_drain(BlockDriverState *bs);
426 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
427 void bdrv_drain_all_begin(void);
428 void bdrv_drain_all_end(void);
429 void bdrv_drain_all(void);
430 
431 #define BDRV_POLL_WHILE(bs, cond) ({                       \
432     BlockDriverState *bs_ = (bs);                          \
433     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
434                    cond); })
435 
436 int bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
437 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
438 int bdrv_has_zero_init_1(BlockDriverState *bs);
439 int bdrv_has_zero_init(BlockDriverState *bs);
440 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
441 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
442 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
443                       int64_t bytes, int64_t *pnum, int64_t *map,
444                       BlockDriverState **file);
445 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
446                             int64_t offset, int64_t bytes, int64_t *pnum,
447                             int64_t *map, BlockDriverState **file);
448 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
449                       int64_t *pnum);
450 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
451                             bool include_base, int64_t offset, int64_t bytes,
452                             int64_t *pnum);
453 
454 bool bdrv_is_read_only(BlockDriverState *bs);
455 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
456                            bool ignore_allow_rdw, Error **errp);
457 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
458                               Error **errp);
459 bool bdrv_is_writable(BlockDriverState *bs);
460 bool bdrv_is_sg(BlockDriverState *bs);
461 bool bdrv_is_inserted(BlockDriverState *bs);
462 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
463 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
464 const char *bdrv_get_format_name(BlockDriverState *bs);
465 BlockDriverState *bdrv_find_node(const char *node_name);
466 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
467 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
468 BlockDriverState *bdrv_lookup_bs(const char *device,
469                                  const char *node_name,
470                                  Error **errp);
471 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
472 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
473 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
474 
475 typedef struct BdrvNextIterator {
476     enum {
477         BDRV_NEXT_BACKEND_ROOTS,
478         BDRV_NEXT_MONITOR_OWNED,
479     } phase;
480     BlockBackend *blk;
481     BlockDriverState *bs;
482 } BdrvNextIterator;
483 
484 BlockDriverState *bdrv_first(BdrvNextIterator *it);
485 BlockDriverState *bdrv_next(BdrvNextIterator *it);
486 void bdrv_next_cleanup(BdrvNextIterator *it);
487 
488 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
489 bool bdrv_is_encrypted(BlockDriverState *bs);
490 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
491                          void *opaque, bool read_only);
492 const char *bdrv_get_node_name(const BlockDriverState *bs);
493 const char *bdrv_get_device_name(const BlockDriverState *bs);
494 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
495 int bdrv_get_flags(BlockDriverState *bs);
496 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
497 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
498                                           Error **errp);
499 void bdrv_round_to_clusters(BlockDriverState *bs,
500                             int64_t offset, int64_t bytes,
501                             int64_t *cluster_offset,
502                             int64_t *cluster_bytes);
503 
504 void bdrv_get_backing_filename(BlockDriverState *bs,
505                                char *filename, int filename_size);
506 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
507 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
508                                                    const char *backing,
509                                                    Error **errp);
510 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
511 
512 int path_has_protocol(const char *path);
513 int path_is_absolute(const char *path);
514 char *path_combine(const char *base_path, const char *filename);
515 
516 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
517 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
518 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
519                       int64_t pos, int size);
520 
521 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
522                       int64_t pos, int size);
523 
524 void bdrv_img_create(const char *filename, const char *fmt,
525                      const char *base_filename, const char *base_fmt,
526                      char *options, uint64_t img_size, int flags,
527                      bool quiet, Error **errp);
528 
529 /* Returns the alignment in bytes that is required so that no bounce buffer
530  * is required throughout the stack */
531 size_t bdrv_min_mem_align(BlockDriverState *bs);
532 /* Returns optimal alignment in bytes for bounce buffer */
533 size_t bdrv_opt_mem_align(BlockDriverState *bs);
534 void *qemu_blockalign(BlockDriverState *bs, size_t size);
535 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
536 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
537 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
538 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
539 
540 void bdrv_enable_copy_on_read(BlockDriverState *bs);
541 void bdrv_disable_copy_on_read(BlockDriverState *bs);
542 
543 void bdrv_ref(BlockDriverState *bs);
544 void bdrv_unref(BlockDriverState *bs);
545 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
546 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
547                              BlockDriverState *child_bs,
548                              const char *child_name,
549                              const BdrvChildRole *child_role,
550                              Error **errp);
551 
552 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
553 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
554 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
555 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
556 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
557 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
558 
559 #define BLKDBG_EVENT(child, evt) \
560     do { \
561         if (child) { \
562             bdrv_debug_event(child->bs, evt); \
563         } \
564     } while (0)
565 
566 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
567 
568 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
569                            const char *tag);
570 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
571 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
572 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
573 
574 /**
575  * bdrv_get_aio_context:
576  *
577  * Returns: the currently bound #AioContext
578  */
579 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
580 
581 /**
582  * Transfer control to @co in the aio context of @bs
583  */
584 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
585 
586 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
587                                  AioContext *new_context, GSList **ignore);
588 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
589                              Error **errp);
590 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
591                                    BdrvChild *ignore_child, Error **errp);
592 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
593                                     GSList **ignore, Error **errp);
594 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
595                               GSList **ignore, Error **errp);
596 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
597 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
598 
599 void bdrv_io_plug(BlockDriverState *bs);
600 void bdrv_io_unplug(BlockDriverState *bs);
601 
602 /**
603  * bdrv_parent_drained_begin_single:
604  *
605  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
606  * any pending activity to cease.
607  */
608 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
609 
610 /**
611  * bdrv_parent_drained_end_single:
612  *
613  * End a quiesced section for the parent of @c.
614  *
615  * This polls @bs's AioContext until all scheduled sub-drained_ends
616  * have settled, which may result in graph changes.
617  */
618 void bdrv_parent_drained_end_single(BdrvChild *c);
619 
620 /**
621  * bdrv_drain_poll:
622  *
623  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
624  * and if @recursive is true its children as well (used for subtree drain).
625  *
626  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
627  * ignore the drain request because they will be drained separately (used for
628  * drain_all).
629  *
630  * This is part of bdrv_drained_begin.
631  */
632 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
633                      BdrvChild *ignore_parent, bool ignore_bds_parents);
634 
635 /**
636  * bdrv_drained_begin:
637  *
638  * Begin a quiesced section for exclusive access to the BDS, by disabling
639  * external request sources including NBD server and device model. Note that
640  * this doesn't block timers or coroutines from submitting more requests, which
641  * means block_job_pause is still necessary.
642  *
643  * This function can be recursive.
644  */
645 void bdrv_drained_begin(BlockDriverState *bs);
646 
647 /**
648  * bdrv_do_drained_begin_quiesce:
649  *
650  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
651  * running requests to complete.
652  */
653 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
654                                    BdrvChild *parent, bool ignore_bds_parents);
655 
656 /**
657  * Like bdrv_drained_begin, but recursively begins a quiesced section for
658  * exclusive access to all child nodes as well.
659  */
660 void bdrv_subtree_drained_begin(BlockDriverState *bs);
661 
662 /**
663  * bdrv_drained_end:
664  *
665  * End a quiescent section started by bdrv_drained_begin().
666  *
667  * This polls @bs's AioContext until all scheduled sub-drained_ends
668  * have settled.  On one hand, that may result in graph changes.  On
669  * the other, this requires that the caller either runs in the main
670  * loop; or that all involved nodes (@bs and all of its parents) are
671  * in the caller's AioContext.
672  */
673 void bdrv_drained_end(BlockDriverState *bs);
674 
675 /**
676  * bdrv_drained_end_no_poll:
677  *
678  * Same as bdrv_drained_end(), but do not poll for the subgraph to
679  * actually become unquiesced.  Therefore, no graph changes will occur
680  * with this function.
681  *
682  * *drained_end_counter is incremented for every background operation
683  * that is scheduled, and will be decremented for every operation once
684  * it settles.  The caller must poll until it reaches 0.  The counter
685  * should be accessed using atomic operations only.
686  */
687 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
688 
689 /**
690  * End a quiescent section started by bdrv_subtree_drained_begin().
691  */
692 void bdrv_subtree_drained_end(BlockDriverState *bs);
693 
694 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
695                     Error **errp);
696 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
697 
698 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
699                                      uint32_t granularity, Error **errp);
700 /**
701  *
702  * bdrv_register_buf/bdrv_unregister_buf:
703  *
704  * Register/unregister a buffer for I/O. For example, VFIO drivers are
705  * interested to know the memory areas that would later be used for I/O, so
706  * that they can prepare IOMMU mapping etc., to get better performance.
707  */
708 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
709 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
710 
711 /**
712  *
713  * bdrv_co_copy_range:
714  *
715  * Do offloaded copy between two children. If the operation is not implemented
716  * by the driver, or if the backend storage doesn't support it, a negative
717  * error code will be returned.
718  *
719  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
720  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
721  * calling copy_file_range(2)) after the first error, thus it should fall back
722  * to a read+write path in the caller level.
723  *
724  * @src: Source child to copy data from
725  * @src_offset: offset in @src image to read data
726  * @dst: Destination child to copy data to
727  * @dst_offset: offset in @dst image to write data
728  * @bytes: number of bytes to copy
729  * @flags: request flags. Supported flags:
730  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
731  *                               write on @dst as if bdrv_co_pwrite_zeroes is
732  *                               called. Used to simplify caller code, or
733  *                               during BlockDriver.bdrv_co_copy_range_from()
734  *                               recursion.
735  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
736  *                                   requests currently in flight.
737  *
738  * Returns: 0 if succeeded; negative error code if failed.
739  **/
740 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
741                                     BdrvChild *dst, uint64_t dst_offset,
742                                     uint64_t bytes, BdrvRequestFlags read_flags,
743                                     BdrvRequestFlags write_flags);
744 #endif
745