xref: /openbmc/qemu/include/block/block.h (revision 438c78da)
1 #ifndef BLOCK_H
2 #define BLOCK_H
3 
4 #include "block/aio.h"
5 #include "qapi/qapi-types-block-core.h"
6 #include "block/aio-wait.h"
7 #include "qemu/iov.h"
8 #include "qemu/coroutine.h"
9 #include "block/accounting.h"
10 #include "block/dirty-bitmap.h"
11 #include "block/blockjob.h"
12 #include "qemu/hbitmap.h"
13 
14 /* block.c */
15 typedef struct BlockDriver BlockDriver;
16 typedef struct BdrvChild BdrvChild;
17 typedef struct BdrvChildRole BdrvChildRole;
18 
19 typedef struct BlockDriverInfo {
20     /* in bytes, 0 if irrelevant */
21     int cluster_size;
22     /* offset at which the VM state can be saved (0 if not possible) */
23     int64_t vm_state_offset;
24     bool is_dirty;
25     /*
26      * True if unallocated blocks read back as zeroes. This is equivalent
27      * to the LBPRZ flag in the SCSI logical block provisioning page.
28      */
29     bool unallocated_blocks_are_zero;
30     /*
31      * True if this block driver only supports compressed writes
32      */
33     bool needs_compressed_writes;
34 } BlockDriverInfo;
35 
36 typedef struct BlockFragInfo {
37     uint64_t allocated_clusters;
38     uint64_t total_clusters;
39     uint64_t fragmented_clusters;
40     uint64_t compressed_clusters;
41 } BlockFragInfo;
42 
43 typedef enum {
44     BDRV_REQ_COPY_ON_READ       = 0x1,
45     BDRV_REQ_ZERO_WRITE         = 0x2,
46 
47     /*
48      * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
49      * that the block driver should unmap (discard) blocks if it is guaranteed
50      * that the result will read back as zeroes. The flag is only passed to the
51      * driver if the block device is opened with BDRV_O_UNMAP.
52      */
53     BDRV_REQ_MAY_UNMAP          = 0x4,
54 
55     /*
56      * The BDRV_REQ_NO_SERIALISING flag is only valid for reads and means that
57      * we don't want wait_serialising_requests() during the read operation.
58      *
59      * This flag is used for backup copy-on-write operations, when we need to
60      * read old data before write (write notifier triggered). It is okay since
61      * we already waited for other serializing requests in the initiating write
62      * (see bdrv_aligned_pwritev), and it is necessary if the initiating write
63      * is already serializing (without the flag, the read would deadlock
64      * waiting for the serialising write to complete).
65      */
66     BDRV_REQ_NO_SERIALISING     = 0x8,
67     BDRV_REQ_FUA                = 0x10,
68     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
69 
70     /* Signifies that this write request will not change the visible disk
71      * content. */
72     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
73 
74     /*
75      * BDRV_REQ_SERIALISING forces request serialisation for writes.
76      * It is used to ensure that writes to the backing file of a backup process
77      * target cannot race with a read of the backup target that defers to the
78      * backing file.
79      *
80      * Note, that BDRV_REQ_SERIALISING is _not_ opposite in meaning to
81      * BDRV_REQ_NO_SERIALISING. A more descriptive name for the latter might be
82      * _DO_NOT_WAIT_FOR_SERIALISING, except that is too long.
83      */
84     BDRV_REQ_SERIALISING        = 0x80,
85 
86     /* Mask of valid flags */
87     BDRV_REQ_MASK               = 0xff,
88 } BdrvRequestFlags;
89 
90 typedef struct BlockSizes {
91     uint32_t phys;
92     uint32_t log;
93 } BlockSizes;
94 
95 typedef struct HDGeometry {
96     uint32_t heads;
97     uint32_t sectors;
98     uint32_t cylinders;
99 } HDGeometry;
100 
101 #define BDRV_O_RDWR        0x0002
102 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
103 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
104 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
105 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
106 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
107 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
108 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
109 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
110 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
111 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
112 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
113 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
114 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
115                                       select an appropriate protocol driver,
116                                       ignoring the format layer */
117 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
118 
119 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
120 
121 
122 /* Option names of options parsed by the block layer */
123 
124 #define BDRV_OPT_CACHE_WB       "cache.writeback"
125 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
126 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
127 #define BDRV_OPT_READ_ONLY      "read-only"
128 #define BDRV_OPT_DISCARD        "discard"
129 #define BDRV_OPT_FORCE_SHARE    "force-share"
130 
131 
132 #define BDRV_SECTOR_BITS   9
133 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
134 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
135 
136 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
137                                      INT_MAX >> BDRV_SECTOR_BITS)
138 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
139 
140 /*
141  * Allocation status flags for bdrv_block_status() and friends.
142  *
143  * Public flags:
144  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
145  * BDRV_BLOCK_ZERO: offset reads as zero
146  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
147  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
148  *                       layer rather than any backing, set by block layer
149  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
150  *                 layer, set by block layer
151  *
152  * Internal flag:
153  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
154  *                 that the block layer recompute the answer from the returned
155  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
156  *
157  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
158  * host offset within the returned BDS that is allocated for the
159  * corresponding raw guest data.  However, whether that offset
160  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
161  *
162  * DATA ZERO OFFSET_VALID
163  *  t    t        t       sectors read as zero, returned file is zero at offset
164  *  t    f        t       sectors read as valid from file at offset
165  *  f    t        t       sectors preallocated, read as zero, returned file not
166  *                        necessarily zero at offset
167  *  f    f        t       sectors preallocated but read from backing_hd,
168  *                        returned file contains garbage at offset
169  *  t    t        f       sectors preallocated, read as zero, unknown offset
170  *  t    f        f       sectors read from unknown file or offset
171  *  f    t        f       not allocated or unknown offset, read as zero
172  *  f    f        f       not allocated or unknown offset, read from backing_hd
173  */
174 #define BDRV_BLOCK_DATA         0x01
175 #define BDRV_BLOCK_ZERO         0x02
176 #define BDRV_BLOCK_OFFSET_VALID 0x04
177 #define BDRV_BLOCK_RAW          0x08
178 #define BDRV_BLOCK_ALLOCATED    0x10
179 #define BDRV_BLOCK_EOF          0x20
180 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
181 
182 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
183 
184 typedef struct BDRVReopenState {
185     BlockDriverState *bs;
186     int flags;
187     BlockdevDetectZeroesOptions detect_zeroes;
188     uint64_t perm, shared_perm;
189     QDict *options;
190     QDict *explicit_options;
191     void *opaque;
192 } BDRVReopenState;
193 
194 /*
195  * Block operation types
196  */
197 typedef enum BlockOpType {
198     BLOCK_OP_TYPE_BACKUP_SOURCE,
199     BLOCK_OP_TYPE_BACKUP_TARGET,
200     BLOCK_OP_TYPE_CHANGE,
201     BLOCK_OP_TYPE_COMMIT_SOURCE,
202     BLOCK_OP_TYPE_COMMIT_TARGET,
203     BLOCK_OP_TYPE_DATAPLANE,
204     BLOCK_OP_TYPE_DRIVE_DEL,
205     BLOCK_OP_TYPE_EJECT,
206     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
207     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
208     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
209     BLOCK_OP_TYPE_MIRROR_SOURCE,
210     BLOCK_OP_TYPE_MIRROR_TARGET,
211     BLOCK_OP_TYPE_RESIZE,
212     BLOCK_OP_TYPE_STREAM,
213     BLOCK_OP_TYPE_REPLACE,
214     BLOCK_OP_TYPE_MAX,
215 } BlockOpType;
216 
217 /* Block node permission constants */
218 enum {
219     /**
220      * A user that has the "permission" of consistent reads is guaranteed that
221      * their view of the contents of the block device is complete and
222      * self-consistent, representing the contents of a disk at a specific
223      * point.
224      *
225      * For most block devices (including their backing files) this is true, but
226      * the property cannot be maintained in a few situations like for
227      * intermediate nodes of a commit block job.
228      */
229     BLK_PERM_CONSISTENT_READ    = 0x01,
230 
231     /** This permission is required to change the visible disk contents. */
232     BLK_PERM_WRITE              = 0x02,
233 
234     /**
235      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
236      * required for writes to the block node when the caller promises that
237      * the visible disk content doesn't change.
238      *
239      * As the BLK_PERM_WRITE permission is strictly stronger, either is
240      * sufficient to perform an unchanging write.
241      */
242     BLK_PERM_WRITE_UNCHANGED    = 0x04,
243 
244     /** This permission is required to change the size of a block node. */
245     BLK_PERM_RESIZE             = 0x08,
246 
247     /**
248      * This permission is required to change the node that this BdrvChild
249      * points to.
250      */
251     BLK_PERM_GRAPH_MOD          = 0x10,
252 
253     BLK_PERM_ALL                = 0x1f,
254 
255     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
256                                  | BLK_PERM_WRITE
257                                  | BLK_PERM_WRITE_UNCHANGED
258                                  | BLK_PERM_RESIZE,
259 
260     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
261 };
262 
263 char *bdrv_perm_names(uint64_t perm);
264 
265 /* disk I/O throttling */
266 void bdrv_init(void);
267 void bdrv_init_with_whitelist(void);
268 bool bdrv_uses_whitelist(void);
269 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
270 BlockDriver *bdrv_find_protocol(const char *filename,
271                                 bool allow_protocol_prefix,
272                                 Error **errp);
273 BlockDriver *bdrv_find_format(const char *format_name);
274 int bdrv_create(BlockDriver *drv, const char* filename,
275                 QemuOpts *opts, Error **errp);
276 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
277 BlockDriverState *bdrv_new(void);
278 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
279                  Error **errp);
280 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
281                        Error **errp);
282 
283 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
284 int bdrv_parse_discard_flags(const char *mode, int *flags);
285 BdrvChild *bdrv_open_child(const char *filename,
286                            QDict *options, const char *bdref_key,
287                            BlockDriverState* parent,
288                            const BdrvChildRole *child_role,
289                            bool allow_none, Error **errp);
290 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
291 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
292                          Error **errp);
293 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
294                            const char *bdref_key, Error **errp);
295 BlockDriverState *bdrv_open(const char *filename, const char *reference,
296                             QDict *options, int flags, Error **errp);
297 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
298                                        int flags, Error **errp);
299 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
300                                     BlockDriverState *bs,
301                                     QDict *options, int flags);
302 int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp);
303 int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp);
304 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
305                         BlockReopenQueue *queue, Error **errp);
306 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
307 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
308 int bdrv_read(BdrvChild *child, int64_t sector_num,
309               uint8_t *buf, int nb_sectors);
310 int bdrv_write(BdrvChild *child, int64_t sector_num,
311                const uint8_t *buf, int nb_sectors);
312 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
313                        int bytes, BdrvRequestFlags flags);
314 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
315 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
316 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
317 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
318 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
319 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
320                      const void *buf, int count);
321 /*
322  * Efficiently zero a region of the disk image.  Note that this is a regular
323  * I/O request like read or write and should have a reasonable size.  This
324  * function is not suitable for zeroing the entire image in a single request
325  * because it may allocate memory for the entire region.
326  */
327 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
328                                        int bytes, BdrvRequestFlags flags);
329 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
330     const char *backing_file);
331 void bdrv_refresh_filename(BlockDriverState *bs);
332 
333 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset,
334                                   PreallocMode prealloc, Error **errp);
335 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
336                   Error **errp);
337 
338 int64_t bdrv_nb_sectors(BlockDriverState *bs);
339 int64_t bdrv_getlength(BlockDriverState *bs);
340 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
341 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
342                                BlockDriverState *in_bs, Error **errp);
343 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
344 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
345 int bdrv_commit(BlockDriverState *bs);
346 int bdrv_change_backing_file(BlockDriverState *bs,
347     const char *backing_file, const char *backing_fmt);
348 void bdrv_register(BlockDriver *bdrv);
349 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
350                            const char *backing_file_str);
351 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
352                                     BlockDriverState *bs);
353 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
354 
355 
356 typedef struct BdrvCheckResult {
357     int corruptions;
358     int leaks;
359     int check_errors;
360     int corruptions_fixed;
361     int leaks_fixed;
362     int64_t image_end_offset;
363     BlockFragInfo bfi;
364 } BdrvCheckResult;
365 
366 typedef enum {
367     BDRV_FIX_LEAKS    = 1,
368     BDRV_FIX_ERRORS   = 2,
369 } BdrvCheckMode;
370 
371 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
372 
373 /* The units of offset and total_work_size may be chosen arbitrarily by the
374  * block driver; total_work_size may change during the course of the amendment
375  * operation */
376 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
377                                       int64_t total_work_size, void *opaque);
378 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
379                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
380                        Error **errp);
381 
382 /* external snapshots */
383 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
384                                       BlockDriverState *candidate);
385 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
386 
387 /* check if a named node can be replaced when doing drive-mirror */
388 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
389                                         const char *node_name, Error **errp);
390 
391 /* async block I/O */
392 void bdrv_aio_cancel(BlockAIOCB *acb);
393 void bdrv_aio_cancel_async(BlockAIOCB *acb);
394 
395 /* sg packet commands */
396 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
397 
398 /* Invalidate any cached metadata used by image formats */
399 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
400 void bdrv_invalidate_cache_all(Error **errp);
401 int bdrv_inactivate_all(void);
402 
403 /* Ensure contents are flushed to disk.  */
404 int bdrv_flush(BlockDriverState *bs);
405 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
406 int bdrv_flush_all(void);
407 void bdrv_close_all(void);
408 void bdrv_drain(BlockDriverState *bs);
409 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
410 void bdrv_drain_all_begin(void);
411 void bdrv_drain_all_end(void);
412 void bdrv_drain_all(void);
413 
414 #define BDRV_POLL_WHILE(bs, cond) ({                       \
415     BlockDriverState *bs_ = (bs);                          \
416     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
417                    cond); })
418 
419 int bdrv_pdiscard(BdrvChild *child, int64_t offset, int bytes);
420 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int bytes);
421 int bdrv_has_zero_init_1(BlockDriverState *bs);
422 int bdrv_has_zero_init(BlockDriverState *bs);
423 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
424 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
425 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
426                       int64_t bytes, int64_t *pnum, int64_t *map,
427                       BlockDriverState **file);
428 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
429                             int64_t offset, int64_t bytes, int64_t *pnum,
430                             int64_t *map, BlockDriverState **file);
431 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
432                       int64_t *pnum);
433 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
434                             int64_t offset, int64_t bytes, int64_t *pnum);
435 
436 bool bdrv_is_read_only(BlockDriverState *bs);
437 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
438                            bool ignore_allow_rdw, Error **errp);
439 int bdrv_set_read_only(BlockDriverState *bs, bool read_only, Error **errp);
440 bool bdrv_is_writable(BlockDriverState *bs);
441 bool bdrv_is_sg(BlockDriverState *bs);
442 bool bdrv_is_inserted(BlockDriverState *bs);
443 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
444 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
445 const char *bdrv_get_format_name(BlockDriverState *bs);
446 BlockDriverState *bdrv_find_node(const char *node_name);
447 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
448 BlockDriverState *bdrv_lookup_bs(const char *device,
449                                  const char *node_name,
450                                  Error **errp);
451 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
452 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
453 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
454 
455 typedef struct BdrvNextIterator {
456     enum {
457         BDRV_NEXT_BACKEND_ROOTS,
458         BDRV_NEXT_MONITOR_OWNED,
459     } phase;
460     BlockBackend *blk;
461     BlockDriverState *bs;
462 } BdrvNextIterator;
463 
464 BlockDriverState *bdrv_first(BdrvNextIterator *it);
465 BlockDriverState *bdrv_next(BdrvNextIterator *it);
466 void bdrv_next_cleanup(BdrvNextIterator *it);
467 
468 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
469 bool bdrv_is_encrypted(BlockDriverState *bs);
470 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
471                          void *opaque);
472 const char *bdrv_get_node_name(const BlockDriverState *bs);
473 const char *bdrv_get_device_name(const BlockDriverState *bs);
474 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
475 int bdrv_get_flags(BlockDriverState *bs);
476 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
477 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs);
478 void bdrv_round_to_clusters(BlockDriverState *bs,
479                             int64_t offset, int64_t bytes,
480                             int64_t *cluster_offset,
481                             int64_t *cluster_bytes);
482 
483 const char *bdrv_get_encrypted_filename(BlockDriverState *bs);
484 void bdrv_get_backing_filename(BlockDriverState *bs,
485                                char *filename, int filename_size);
486 void bdrv_get_full_backing_filename(BlockDriverState *bs,
487                                     char *dest, size_t sz, Error **errp);
488 void bdrv_get_full_backing_filename_from_filename(const char *backed,
489                                                   const char *backing,
490                                                   char *dest, size_t sz,
491                                                   Error **errp);
492 
493 int path_has_protocol(const char *path);
494 int path_is_absolute(const char *path);
495 void path_combine(char *dest, int dest_size,
496                   const char *base_path,
497                   const char *filename);
498 
499 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
500 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
501 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
502                       int64_t pos, int size);
503 
504 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
505                       int64_t pos, int size);
506 
507 void bdrv_img_create(const char *filename, const char *fmt,
508                      const char *base_filename, const char *base_fmt,
509                      char *options, uint64_t img_size, int flags,
510                      bool quiet, Error **errp);
511 
512 /* Returns the alignment in bytes that is required so that no bounce buffer
513  * is required throughout the stack */
514 size_t bdrv_min_mem_align(BlockDriverState *bs);
515 /* Returns optimal alignment in bytes for bounce buffer */
516 size_t bdrv_opt_mem_align(BlockDriverState *bs);
517 void *qemu_blockalign(BlockDriverState *bs, size_t size);
518 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
519 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
520 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
521 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
522 
523 void bdrv_enable_copy_on_read(BlockDriverState *bs);
524 void bdrv_disable_copy_on_read(BlockDriverState *bs);
525 
526 void bdrv_ref(BlockDriverState *bs);
527 void bdrv_unref(BlockDriverState *bs);
528 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
529 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
530                              BlockDriverState *child_bs,
531                              const char *child_name,
532                              const BdrvChildRole *child_role,
533                              Error **errp);
534 
535 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
536 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
537 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
538 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
539 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
540 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
541 
542 #define BLKDBG_EVENT(child, evt) \
543     do { \
544         if (child) { \
545             bdrv_debug_event(child->bs, evt); \
546         } \
547     } while (0)
548 
549 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
550 
551 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
552                            const char *tag);
553 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
554 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
555 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
556 
557 /**
558  * bdrv_get_aio_context:
559  *
560  * Returns: the currently bound #AioContext
561  */
562 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
563 
564 /**
565  * Transfer control to @co in the aio context of @bs
566  */
567 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
568 
569 /**
570  * bdrv_set_aio_context:
571  *
572  * Changes the #AioContext used for fd handlers, timers, and BHs by this
573  * BlockDriverState and all its children.
574  *
575  * This function must be called with iothread lock held.
576  */
577 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
578 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
579 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
580 
581 void bdrv_io_plug(BlockDriverState *bs);
582 void bdrv_io_unplug(BlockDriverState *bs);
583 
584 /**
585  * bdrv_parent_drained_begin:
586  *
587  * Begin a quiesced section of all users of @bs. This is part of
588  * bdrv_drained_begin.
589  */
590 void bdrv_parent_drained_begin(BlockDriverState *bs, BdrvChild *ignore,
591                                bool ignore_bds_parents);
592 
593 /**
594  * bdrv_parent_drained_begin_single:
595  *
596  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
597  * any pending activity to cease.
598  */
599 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
600 
601 /**
602  * bdrv_parent_drained_end:
603  *
604  * End a quiesced section of all users of @bs. This is part of
605  * bdrv_drained_end.
606  */
607 void bdrv_parent_drained_end(BlockDriverState *bs, BdrvChild *ignore,
608                              bool ignore_bds_parents);
609 
610 /**
611  * bdrv_drain_poll:
612  *
613  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
614  * and if @recursive is true its children as well (used for subtree drain).
615  *
616  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
617  * ignore the drain request because they will be drained separately (used for
618  * drain_all).
619  *
620  * This is part of bdrv_drained_begin.
621  */
622 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
623                      BdrvChild *ignore_parent, bool ignore_bds_parents);
624 
625 /**
626  * bdrv_drained_begin:
627  *
628  * Begin a quiesced section for exclusive access to the BDS, by disabling
629  * external request sources including NBD server and device model. Note that
630  * this doesn't block timers or coroutines from submitting more requests, which
631  * means block_job_pause is still necessary.
632  *
633  * This function can be recursive.
634  */
635 void bdrv_drained_begin(BlockDriverState *bs);
636 
637 /**
638  * bdrv_do_drained_begin_quiesce:
639  *
640  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
641  * running requests to complete.
642  */
643 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
644                                    BdrvChild *parent, bool ignore_bds_parents);
645 
646 /**
647  * Like bdrv_drained_begin, but recursively begins a quiesced section for
648  * exclusive access to all child nodes as well.
649  */
650 void bdrv_subtree_drained_begin(BlockDriverState *bs);
651 
652 /**
653  * bdrv_drained_end:
654  *
655  * End a quiescent section started by bdrv_drained_begin().
656  */
657 void bdrv_drained_end(BlockDriverState *bs);
658 
659 /**
660  * End a quiescent section started by bdrv_subtree_drained_begin().
661  */
662 void bdrv_subtree_drained_end(BlockDriverState *bs);
663 
664 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
665                     Error **errp);
666 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
667 
668 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
669                                      uint32_t granularity, Error **errp);
670 /**
671  *
672  * bdrv_register_buf/bdrv_unregister_buf:
673  *
674  * Register/unregister a buffer for I/O. For example, VFIO drivers are
675  * interested to know the memory areas that would later be used for I/O, so
676  * that they can prepare IOMMU mapping etc., to get better performance.
677  */
678 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
679 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
680 
681 /**
682  *
683  * bdrv_co_copy_range:
684  *
685  * Do offloaded copy between two children. If the operation is not implemented
686  * by the driver, or if the backend storage doesn't support it, a negative
687  * error code will be returned.
688  *
689  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
690  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
691  * calling copy_file_range(2)) after the first error, thus it should fall back
692  * to a read+write path in the caller level.
693  *
694  * @src: Source child to copy data from
695  * @src_offset: offset in @src image to read data
696  * @dst: Destination child to copy data to
697  * @dst_offset: offset in @dst image to write data
698  * @bytes: number of bytes to copy
699  * @flags: request flags. Supported flags:
700  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
701  *                               write on @dst as if bdrv_co_pwrite_zeroes is
702  *                               called. Used to simplify caller code, or
703  *                               during BlockDriver.bdrv_co_copy_range_from()
704  *                               recursion.
705  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
706  *                                   requests currently in flight.
707  *
708  * Returns: 0 if succeeded; negative error code if failed.
709  **/
710 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
711                                     BdrvChild *dst, uint64_t dst_offset,
712                                     uint64_t bytes, BdrvRequestFlags read_flags,
713                                     BdrvRequestFlags write_flags);
714 #endif
715