xref: /openbmc/qemu/include/block/block.h (revision 2e5b09fd)
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 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
147 
148 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
149                                      INT_MAX >> BDRV_SECTOR_BITS)
150 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
151 
152 /*
153  * Allocation status flags for bdrv_block_status() and friends.
154  *
155  * Public flags:
156  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
157  * BDRV_BLOCK_ZERO: offset reads as zero
158  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
159  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
160  *                       layer rather than any backing, set by block layer
161  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
162  *                 layer, set by block layer
163  *
164  * Internal flags:
165  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
166  *                 that the block layer recompute the answer from the returned
167  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
168  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
169  *                     zeroes in file child of current block node inside
170  *                     returned region. Only valid together with both
171  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
172  *                     appear with BDRV_BLOCK_ZERO.
173  *
174  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
175  * host offset within the returned BDS that is allocated for the
176  * corresponding raw guest data.  However, whether that offset
177  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
178  *
179  * DATA ZERO OFFSET_VALID
180  *  t    t        t       sectors read as zero, returned file is zero at offset
181  *  t    f        t       sectors read as valid from file at offset
182  *  f    t        t       sectors preallocated, read as zero, returned file not
183  *                        necessarily zero at offset
184  *  f    f        t       sectors preallocated but read from backing_hd,
185  *                        returned file contains garbage at offset
186  *  t    t        f       sectors preallocated, read as zero, unknown offset
187  *  t    f        f       sectors read from unknown file or offset
188  *  f    t        f       not allocated or unknown offset, read as zero
189  *  f    f        f       not allocated or unknown offset, read from backing_hd
190  */
191 #define BDRV_BLOCK_DATA         0x01
192 #define BDRV_BLOCK_ZERO         0x02
193 #define BDRV_BLOCK_OFFSET_VALID 0x04
194 #define BDRV_BLOCK_RAW          0x08
195 #define BDRV_BLOCK_ALLOCATED    0x10
196 #define BDRV_BLOCK_EOF          0x20
197 #define BDRV_BLOCK_RECURSE      0x40
198 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
199 
200 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
201 
202 typedef struct BDRVReopenState {
203     BlockDriverState *bs;
204     int flags;
205     BlockdevDetectZeroesOptions detect_zeroes;
206     bool backing_missing;
207     bool replace_backing_bs;  /* new_backing_bs is ignored if this is false */
208     BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */
209     uint64_t perm, shared_perm;
210     QDict *options;
211     QDict *explicit_options;
212     void *opaque;
213 } BDRVReopenState;
214 
215 /*
216  * Block operation types
217  */
218 typedef enum BlockOpType {
219     BLOCK_OP_TYPE_BACKUP_SOURCE,
220     BLOCK_OP_TYPE_BACKUP_TARGET,
221     BLOCK_OP_TYPE_CHANGE,
222     BLOCK_OP_TYPE_COMMIT_SOURCE,
223     BLOCK_OP_TYPE_COMMIT_TARGET,
224     BLOCK_OP_TYPE_DATAPLANE,
225     BLOCK_OP_TYPE_DRIVE_DEL,
226     BLOCK_OP_TYPE_EJECT,
227     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
228     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
229     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
230     BLOCK_OP_TYPE_MIRROR_SOURCE,
231     BLOCK_OP_TYPE_MIRROR_TARGET,
232     BLOCK_OP_TYPE_RESIZE,
233     BLOCK_OP_TYPE_STREAM,
234     BLOCK_OP_TYPE_REPLACE,
235     BLOCK_OP_TYPE_MAX,
236 } BlockOpType;
237 
238 /* Block node permission constants */
239 enum {
240     /**
241      * A user that has the "permission" of consistent reads is guaranteed that
242      * their view of the contents of the block device is complete and
243      * self-consistent, representing the contents of a disk at a specific
244      * point.
245      *
246      * For most block devices (including their backing files) this is true, but
247      * the property cannot be maintained in a few situations like for
248      * intermediate nodes of a commit block job.
249      */
250     BLK_PERM_CONSISTENT_READ    = 0x01,
251 
252     /** This permission is required to change the visible disk contents. */
253     BLK_PERM_WRITE              = 0x02,
254 
255     /**
256      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
257      * required for writes to the block node when the caller promises that
258      * the visible disk content doesn't change.
259      *
260      * As the BLK_PERM_WRITE permission is strictly stronger, either is
261      * sufficient to perform an unchanging write.
262      */
263     BLK_PERM_WRITE_UNCHANGED    = 0x04,
264 
265     /** This permission is required to change the size of a block node. */
266     BLK_PERM_RESIZE             = 0x08,
267 
268     /**
269      * This permission is required to change the node that this BdrvChild
270      * points to.
271      */
272     BLK_PERM_GRAPH_MOD          = 0x10,
273 
274     BLK_PERM_ALL                = 0x1f,
275 
276     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
277                                  | BLK_PERM_WRITE
278                                  | BLK_PERM_WRITE_UNCHANGED
279                                  | BLK_PERM_RESIZE,
280 
281     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
282 };
283 
284 char *bdrv_perm_names(uint64_t perm);
285 
286 /* disk I/O throttling */
287 void bdrv_init(void);
288 void bdrv_init_with_whitelist(void);
289 bool bdrv_uses_whitelist(void);
290 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
291 BlockDriver *bdrv_find_protocol(const char *filename,
292                                 bool allow_protocol_prefix,
293                                 Error **errp);
294 BlockDriver *bdrv_find_format(const char *format_name);
295 int bdrv_create(BlockDriver *drv, const char* filename,
296                 QemuOpts *opts, Error **errp);
297 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
298 BlockDriverState *bdrv_new(void);
299 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
300                  Error **errp);
301 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
302                        Error **errp);
303 
304 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
305 int bdrv_parse_discard_flags(const char *mode, int *flags);
306 BdrvChild *bdrv_open_child(const char *filename,
307                            QDict *options, const char *bdref_key,
308                            BlockDriverState* parent,
309                            const BdrvChildRole *child_role,
310                            bool allow_none, Error **errp);
311 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
312 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
313                          Error **errp);
314 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
315                            const char *bdref_key, Error **errp);
316 BlockDriverState *bdrv_open(const char *filename, const char *reference,
317                             QDict *options, int flags, Error **errp);
318 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
319                                        int flags, Error **errp);
320 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
321                                     BlockDriverState *bs, QDict *options,
322                                     bool keep_old_opts);
323 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
324 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
325                               Error **errp);
326 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
327                         BlockReopenQueue *queue, Error **errp);
328 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
329 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
330 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
331                        int bytes, BdrvRequestFlags flags);
332 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
333 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
334 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
335 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
336 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
337 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
338                      const void *buf, int count);
339 /*
340  * Efficiently zero a region of the disk image.  Note that this is a regular
341  * I/O request like read or write and should have a reasonable size.  This
342  * function is not suitable for zeroing the entire image in a single request
343  * because it may allocate memory for the entire region.
344  */
345 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
346                                        int bytes, BdrvRequestFlags flags);
347 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
348     const char *backing_file);
349 void bdrv_refresh_filename(BlockDriverState *bs);
350 
351 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset,
352                                   PreallocMode prealloc, Error **errp);
353 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
354                   Error **errp);
355 
356 int64_t bdrv_nb_sectors(BlockDriverState *bs);
357 int64_t bdrv_getlength(BlockDriverState *bs);
358 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
359 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
360                                BlockDriverState *in_bs, Error **errp);
361 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
362 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
363 int bdrv_commit(BlockDriverState *bs);
364 int bdrv_change_backing_file(BlockDriverState *bs,
365     const char *backing_file, const char *backing_fmt);
366 void bdrv_register(BlockDriver *bdrv);
367 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
368                            const char *backing_file_str);
369 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
370                                     BlockDriverState *bs);
371 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
372 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
373                                   Error **errp);
374 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
375                               Error **errp);
376 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
377 
378 
379 typedef struct BdrvCheckResult {
380     int corruptions;
381     int leaks;
382     int check_errors;
383     int corruptions_fixed;
384     int leaks_fixed;
385     int64_t image_end_offset;
386     BlockFragInfo bfi;
387 } BdrvCheckResult;
388 
389 typedef enum {
390     BDRV_FIX_LEAKS    = 1,
391     BDRV_FIX_ERRORS   = 2,
392 } BdrvCheckMode;
393 
394 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
395 
396 /* The units of offset and total_work_size may be chosen arbitrarily by the
397  * block driver; total_work_size may change during the course of the amendment
398  * operation */
399 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
400                                       int64_t total_work_size, void *opaque);
401 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
402                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
403                        Error **errp);
404 
405 /* external snapshots */
406 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
407                                       BlockDriverState *candidate);
408 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
409 
410 /* check if a named node can be replaced when doing drive-mirror */
411 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
412                                         const char *node_name, Error **errp);
413 
414 /* async block I/O */
415 void bdrv_aio_cancel(BlockAIOCB *acb);
416 void bdrv_aio_cancel_async(BlockAIOCB *acb);
417 
418 /* sg packet commands */
419 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
420 
421 /* Invalidate any cached metadata used by image formats */
422 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
423 void bdrv_invalidate_cache_all(Error **errp);
424 int bdrv_inactivate_all(void);
425 
426 /* Ensure contents are flushed to disk.  */
427 int bdrv_flush(BlockDriverState *bs);
428 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
429 int bdrv_flush_all(void);
430 void bdrv_close_all(void);
431 void bdrv_drain(BlockDriverState *bs);
432 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
433 void bdrv_drain_all_begin(void);
434 void bdrv_drain_all_end(void);
435 void bdrv_drain_all(void);
436 
437 #define BDRV_POLL_WHILE(bs, cond) ({                       \
438     BlockDriverState *bs_ = (bs);                          \
439     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
440                    cond); })
441 
442 int bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
443 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
444 int bdrv_has_zero_init_1(BlockDriverState *bs);
445 int bdrv_has_zero_init(BlockDriverState *bs);
446 int bdrv_has_zero_init_truncate(BlockDriverState *bs);
447 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
448 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
449 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
450                       int64_t bytes, int64_t *pnum, int64_t *map,
451                       BlockDriverState **file);
452 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
453                             int64_t offset, int64_t bytes, int64_t *pnum,
454                             int64_t *map, BlockDriverState **file);
455 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
456                       int64_t *pnum);
457 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
458                             bool include_base, int64_t offset, int64_t bytes,
459                             int64_t *pnum);
460 
461 bool bdrv_is_read_only(BlockDriverState *bs);
462 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
463                            bool ignore_allow_rdw, Error **errp);
464 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
465                               Error **errp);
466 bool bdrv_is_writable(BlockDriverState *bs);
467 bool bdrv_is_sg(BlockDriverState *bs);
468 bool bdrv_is_inserted(BlockDriverState *bs);
469 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
470 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
471 const char *bdrv_get_format_name(BlockDriverState *bs);
472 BlockDriverState *bdrv_find_node(const char *node_name);
473 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
474 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
475 BlockDriverState *bdrv_lookup_bs(const char *device,
476                                  const char *node_name,
477                                  Error **errp);
478 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
479 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
480 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
481 
482 typedef struct BdrvNextIterator {
483     enum {
484         BDRV_NEXT_BACKEND_ROOTS,
485         BDRV_NEXT_MONITOR_OWNED,
486     } phase;
487     BlockBackend *blk;
488     BlockDriverState *bs;
489 } BdrvNextIterator;
490 
491 BlockDriverState *bdrv_first(BdrvNextIterator *it);
492 BlockDriverState *bdrv_next(BdrvNextIterator *it);
493 void bdrv_next_cleanup(BdrvNextIterator *it);
494 
495 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
496 bool bdrv_is_encrypted(BlockDriverState *bs);
497 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
498                          void *opaque, bool read_only);
499 const char *bdrv_get_node_name(const BlockDriverState *bs);
500 const char *bdrv_get_device_name(const BlockDriverState *bs);
501 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
502 int bdrv_get_flags(BlockDriverState *bs);
503 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
504 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
505                                           Error **errp);
506 void bdrv_round_to_clusters(BlockDriverState *bs,
507                             int64_t offset, int64_t bytes,
508                             int64_t *cluster_offset,
509                             int64_t *cluster_bytes);
510 
511 void bdrv_get_backing_filename(BlockDriverState *bs,
512                                char *filename, int filename_size);
513 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
514 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
515                                                    const char *backing,
516                                                    Error **errp);
517 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
518 
519 int path_has_protocol(const char *path);
520 int path_is_absolute(const char *path);
521 char *path_combine(const char *base_path, const char *filename);
522 
523 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
524 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
525 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
526                       int64_t pos, int size);
527 
528 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
529                       int64_t pos, int size);
530 
531 void bdrv_img_create(const char *filename, const char *fmt,
532                      const char *base_filename, const char *base_fmt,
533                      char *options, uint64_t img_size, int flags,
534                      bool quiet, Error **errp);
535 
536 /* Returns the alignment in bytes that is required so that no bounce buffer
537  * is required throughout the stack */
538 size_t bdrv_min_mem_align(BlockDriverState *bs);
539 /* Returns optimal alignment in bytes for bounce buffer */
540 size_t bdrv_opt_mem_align(BlockDriverState *bs);
541 void *qemu_blockalign(BlockDriverState *bs, size_t size);
542 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
543 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
544 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
545 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
546 
547 void bdrv_enable_copy_on_read(BlockDriverState *bs);
548 void bdrv_disable_copy_on_read(BlockDriverState *bs);
549 
550 void bdrv_ref(BlockDriverState *bs);
551 void bdrv_unref(BlockDriverState *bs);
552 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
553 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
554                              BlockDriverState *child_bs,
555                              const char *child_name,
556                              const BdrvChildRole *child_role,
557                              Error **errp);
558 
559 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
560 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
561 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
562 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
563 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
564 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
565 
566 #define BLKDBG_EVENT(child, evt) \
567     do { \
568         if (child) { \
569             bdrv_debug_event(child->bs, evt); \
570         } \
571     } while (0)
572 
573 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
574 
575 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
576                            const char *tag);
577 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
578 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
579 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
580 
581 /**
582  * bdrv_get_aio_context:
583  *
584  * Returns: the currently bound #AioContext
585  */
586 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
587 
588 /**
589  * Transfer control to @co in the aio context of @bs
590  */
591 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
592 
593 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
594                                  AioContext *new_context, GSList **ignore);
595 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
596                              Error **errp);
597 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
598                                    BdrvChild *ignore_child, Error **errp);
599 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
600                                     GSList **ignore, Error **errp);
601 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
602                               GSList **ignore, Error **errp);
603 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
604 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
605 
606 void bdrv_io_plug(BlockDriverState *bs);
607 void bdrv_io_unplug(BlockDriverState *bs);
608 
609 /**
610  * bdrv_parent_drained_begin_single:
611  *
612  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
613  * any pending activity to cease.
614  */
615 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
616 
617 /**
618  * bdrv_parent_drained_end_single:
619  *
620  * End a quiesced section for the parent of @c.
621  *
622  * This polls @bs's AioContext until all scheduled sub-drained_ends
623  * have settled, which may result in graph changes.
624  */
625 void bdrv_parent_drained_end_single(BdrvChild *c);
626 
627 /**
628  * bdrv_drain_poll:
629  *
630  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
631  * and if @recursive is true its children as well (used for subtree drain).
632  *
633  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
634  * ignore the drain request because they will be drained separately (used for
635  * drain_all).
636  *
637  * This is part of bdrv_drained_begin.
638  */
639 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
640                      BdrvChild *ignore_parent, bool ignore_bds_parents);
641 
642 /**
643  * bdrv_drained_begin:
644  *
645  * Begin a quiesced section for exclusive access to the BDS, by disabling
646  * external request sources including NBD server and device model. Note that
647  * this doesn't block timers or coroutines from submitting more requests, which
648  * means block_job_pause is still necessary.
649  *
650  * This function can be recursive.
651  */
652 void bdrv_drained_begin(BlockDriverState *bs);
653 
654 /**
655  * bdrv_do_drained_begin_quiesce:
656  *
657  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
658  * running requests to complete.
659  */
660 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
661                                    BdrvChild *parent, bool ignore_bds_parents);
662 
663 /**
664  * Like bdrv_drained_begin, but recursively begins a quiesced section for
665  * exclusive access to all child nodes as well.
666  */
667 void bdrv_subtree_drained_begin(BlockDriverState *bs);
668 
669 /**
670  * bdrv_drained_end:
671  *
672  * End a quiescent section started by bdrv_drained_begin().
673  *
674  * This polls @bs's AioContext until all scheduled sub-drained_ends
675  * have settled.  On one hand, that may result in graph changes.  On
676  * the other, this requires that the caller either runs in the main
677  * loop; or that all involved nodes (@bs and all of its parents) are
678  * in the caller's AioContext.
679  */
680 void bdrv_drained_end(BlockDriverState *bs);
681 
682 /**
683  * bdrv_drained_end_no_poll:
684  *
685  * Same as bdrv_drained_end(), but do not poll for the subgraph to
686  * actually become unquiesced.  Therefore, no graph changes will occur
687  * with this function.
688  *
689  * *drained_end_counter is incremented for every background operation
690  * that is scheduled, and will be decremented for every operation once
691  * it settles.  The caller must poll until it reaches 0.  The counter
692  * should be accessed using atomic operations only.
693  */
694 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
695 
696 /**
697  * End a quiescent section started by bdrv_subtree_drained_begin().
698  */
699 void bdrv_subtree_drained_end(BlockDriverState *bs);
700 
701 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
702                     Error **errp);
703 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
704 
705 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
706                                      uint32_t granularity, Error **errp);
707 /**
708  *
709  * bdrv_register_buf/bdrv_unregister_buf:
710  *
711  * Register/unregister a buffer for I/O. For example, VFIO drivers are
712  * interested to know the memory areas that would later be used for I/O, so
713  * that they can prepare IOMMU mapping etc., to get better performance.
714  */
715 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
716 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
717 
718 /**
719  *
720  * bdrv_co_copy_range:
721  *
722  * Do offloaded copy between two children. If the operation is not implemented
723  * by the driver, or if the backend storage doesn't support it, a negative
724  * error code will be returned.
725  *
726  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
727  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
728  * calling copy_file_range(2)) after the first error, thus it should fall back
729  * to a read+write path in the caller level.
730  *
731  * @src: Source child to copy data from
732  * @src_offset: offset in @src image to read data
733  * @dst: Destination child to copy data to
734  * @dst_offset: offset in @dst image to write data
735  * @bytes: number of bytes to copy
736  * @flags: request flags. Supported flags:
737  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
738  *                               write on @dst as if bdrv_co_pwrite_zeroes is
739  *                               called. Used to simplify caller code, or
740  *                               during BlockDriver.bdrv_co_copy_range_from()
741  *                               recursion.
742  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
743  *                                   requests currently in flight.
744  *
745  * Returns: 0 if succeeded; negative error code if failed.
746  **/
747 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
748                                     BdrvChild *dst, uint64_t dst_offset,
749                                     uint64_t bytes, BdrvRequestFlags read_flags,
750                                     BdrvRequestFlags write_flags);
751 #endif
752