xref: /openbmc/qemu/include/block/block.h (revision 6598f0cd)
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     /* The BDRV_REQ_MAY_UNMAP flag is used to indicate that the block driver
47      * is allowed to optimize a write zeroes request by unmapping (discarding)
48      * blocks if it is guaranteed that the result will read back as
49      * zeroes. The flag is only passed to the driver if the block device is
50      * 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     /* Mask of valid flags */
86     BDRV_REQ_MASK               = 0xff,
87 } BdrvRequestFlags;
88 
89 typedef struct BlockSizes {
90     uint32_t phys;
91     uint32_t log;
92 } BlockSizes;
93 
94 typedef struct HDGeometry {
95     uint32_t heads;
96     uint32_t sectors;
97     uint32_t cylinders;
98 } HDGeometry;
99 
100 #define BDRV_O_RDWR        0x0002
101 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
102 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
103 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
104 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
105 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
106 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
107 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
108 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
109 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
110 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
111 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
112 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
113 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
114                                       select an appropriate protocol driver,
115                                       ignoring the format layer */
116 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
117 
118 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
119 
120 
121 /* Option names of options parsed by the block layer */
122 
123 #define BDRV_OPT_CACHE_WB       "cache.writeback"
124 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
125 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
126 #define BDRV_OPT_READ_ONLY      "read-only"
127 #define BDRV_OPT_DISCARD        "discard"
128 #define BDRV_OPT_FORCE_SHARE    "force-share"
129 
130 
131 #define BDRV_SECTOR_BITS   9
132 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
133 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
134 
135 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
136                                      INT_MAX >> BDRV_SECTOR_BITS)
137 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
138 
139 /*
140  * Allocation status flags for bdrv_block_status() and friends.
141  *
142  * Public flags:
143  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
144  * BDRV_BLOCK_ZERO: offset reads as zero
145  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
146  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
147  *                       layer rather than any backing, set by block layer
148  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
149  *                 layer, set by block layer
150  *
151  * Internal flag:
152  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
153  *                 that the block layer recompute the answer from the returned
154  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
155  *
156  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
157  * host offset within the returned BDS that is allocated for the
158  * corresponding raw guest data.  However, whether that offset
159  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
160  *
161  * DATA ZERO OFFSET_VALID
162  *  t    t        t       sectors read as zero, returned file is zero at offset
163  *  t    f        t       sectors read as valid from file at offset
164  *  f    t        t       sectors preallocated, read as zero, returned file not
165  *                        necessarily zero at offset
166  *  f    f        t       sectors preallocated but read from backing_hd,
167  *                        returned file contains garbage at offset
168  *  t    t        f       sectors preallocated, read as zero, unknown offset
169  *  t    f        f       sectors read from unknown file or offset
170  *  f    t        f       not allocated or unknown offset, read as zero
171  *  f    f        f       not allocated or unknown offset, read from backing_hd
172  */
173 #define BDRV_BLOCK_DATA         0x01
174 #define BDRV_BLOCK_ZERO         0x02
175 #define BDRV_BLOCK_OFFSET_VALID 0x04
176 #define BDRV_BLOCK_RAW          0x08
177 #define BDRV_BLOCK_ALLOCATED    0x10
178 #define BDRV_BLOCK_EOF          0x20
179 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
180 
181 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
182 
183 typedef struct BDRVReopenState {
184     BlockDriverState *bs;
185     int flags;
186     uint64_t perm, shared_perm;
187     QDict *options;
188     QDict *explicit_options;
189     void *opaque;
190 } BDRVReopenState;
191 
192 /*
193  * Block operation types
194  */
195 typedef enum BlockOpType {
196     BLOCK_OP_TYPE_BACKUP_SOURCE,
197     BLOCK_OP_TYPE_BACKUP_TARGET,
198     BLOCK_OP_TYPE_CHANGE,
199     BLOCK_OP_TYPE_COMMIT_SOURCE,
200     BLOCK_OP_TYPE_COMMIT_TARGET,
201     BLOCK_OP_TYPE_DATAPLANE,
202     BLOCK_OP_TYPE_DRIVE_DEL,
203     BLOCK_OP_TYPE_EJECT,
204     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
205     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
206     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
207     BLOCK_OP_TYPE_MIRROR_SOURCE,
208     BLOCK_OP_TYPE_MIRROR_TARGET,
209     BLOCK_OP_TYPE_RESIZE,
210     BLOCK_OP_TYPE_STREAM,
211     BLOCK_OP_TYPE_REPLACE,
212     BLOCK_OP_TYPE_MAX,
213 } BlockOpType;
214 
215 /* Block node permission constants */
216 enum {
217     /**
218      * A user that has the "permission" of consistent reads is guaranteed that
219      * their view of the contents of the block device is complete and
220      * self-consistent, representing the contents of a disk at a specific
221      * point.
222      *
223      * For most block devices (including their backing files) this is true, but
224      * the property cannot be maintained in a few situations like for
225      * intermediate nodes of a commit block job.
226      */
227     BLK_PERM_CONSISTENT_READ    = 0x01,
228 
229     /** This permission is required to change the visible disk contents. */
230     BLK_PERM_WRITE              = 0x02,
231 
232     /**
233      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
234      * required for writes to the block node when the caller promises that
235      * the visible disk content doesn't change.
236      *
237      * As the BLK_PERM_WRITE permission is strictly stronger, either is
238      * sufficient to perform an unchanging write.
239      */
240     BLK_PERM_WRITE_UNCHANGED    = 0x04,
241 
242     /** This permission is required to change the size of a block node. */
243     BLK_PERM_RESIZE             = 0x08,
244 
245     /**
246      * This permission is required to change the node that this BdrvChild
247      * points to.
248      */
249     BLK_PERM_GRAPH_MOD          = 0x10,
250 
251     BLK_PERM_ALL                = 0x1f,
252 
253     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
254                                  | BLK_PERM_WRITE
255                                  | BLK_PERM_WRITE_UNCHANGED
256                                  | BLK_PERM_RESIZE,
257 
258     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
259 };
260 
261 char *bdrv_perm_names(uint64_t perm);
262 
263 /* disk I/O throttling */
264 void bdrv_init(void);
265 void bdrv_init_with_whitelist(void);
266 bool bdrv_uses_whitelist(void);
267 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
268 BlockDriver *bdrv_find_protocol(const char *filename,
269                                 bool allow_protocol_prefix,
270                                 Error **errp);
271 BlockDriver *bdrv_find_format(const char *format_name);
272 int bdrv_create(BlockDriver *drv, const char* filename,
273                 QemuOpts *opts, Error **errp);
274 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
275 BlockDriverState *bdrv_new(void);
276 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
277                  Error **errp);
278 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
279                        Error **errp);
280 
281 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
282 int bdrv_parse_discard_flags(const char *mode, int *flags);
283 BdrvChild *bdrv_open_child(const char *filename,
284                            QDict *options, const char *bdref_key,
285                            BlockDriverState* parent,
286                            const BdrvChildRole *child_role,
287                            bool allow_none, Error **errp);
288 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
289 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
290                          Error **errp);
291 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
292                            const char *bdref_key, Error **errp);
293 BlockDriverState *bdrv_open(const char *filename, const char *reference,
294                             QDict *options, int flags, Error **errp);
295 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
296                                        int flags, Error **errp);
297 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
298                                     BlockDriverState *bs,
299                                     QDict *options, int flags);
300 int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp);
301 int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp);
302 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
303                         BlockReopenQueue *queue, Error **errp);
304 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
305 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
306 int bdrv_read(BdrvChild *child, int64_t sector_num,
307               uint8_t *buf, int nb_sectors);
308 int bdrv_write(BdrvChild *child, int64_t sector_num,
309                const uint8_t *buf, int nb_sectors);
310 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
311                        int bytes, BdrvRequestFlags flags);
312 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
313 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
314 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
315 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
316 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
317 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
318                      const void *buf, int count);
319 /*
320  * Efficiently zero a region of the disk image.  Note that this is a regular
321  * I/O request like read or write and should have a reasonable size.  This
322  * function is not suitable for zeroing the entire image in a single request
323  * because it may allocate memory for the entire region.
324  */
325 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
326                                        int bytes, BdrvRequestFlags flags);
327 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
328     const char *backing_file);
329 void bdrv_refresh_filename(BlockDriverState *bs);
330 
331 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset,
332                                   PreallocMode prealloc, Error **errp);
333 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
334                   Error **errp);
335 
336 int64_t bdrv_nb_sectors(BlockDriverState *bs);
337 int64_t bdrv_getlength(BlockDriverState *bs);
338 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
339 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
340                                BlockDriverState *in_bs, Error **errp);
341 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
342 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
343 int bdrv_commit(BlockDriverState *bs);
344 int bdrv_change_backing_file(BlockDriverState *bs,
345     const char *backing_file, const char *backing_fmt);
346 void bdrv_register(BlockDriver *bdrv);
347 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
348                            const char *backing_file_str);
349 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
350                                     BlockDriverState *bs);
351 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
352 
353 
354 typedef struct BdrvCheckResult {
355     int corruptions;
356     int leaks;
357     int check_errors;
358     int corruptions_fixed;
359     int leaks_fixed;
360     int64_t image_end_offset;
361     BlockFragInfo bfi;
362 } BdrvCheckResult;
363 
364 typedef enum {
365     BDRV_FIX_LEAKS    = 1,
366     BDRV_FIX_ERRORS   = 2,
367 } BdrvCheckMode;
368 
369 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
370 
371 /* The units of offset and total_work_size may be chosen arbitrarily by the
372  * block driver; total_work_size may change during the course of the amendment
373  * operation */
374 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
375                                       int64_t total_work_size, void *opaque);
376 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
377                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
378                        Error **errp);
379 
380 /* external snapshots */
381 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
382                                       BlockDriverState *candidate);
383 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
384 
385 /* check if a named node can be replaced when doing drive-mirror */
386 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
387                                         const char *node_name, Error **errp);
388 
389 /* async block I/O */
390 void bdrv_aio_cancel(BlockAIOCB *acb);
391 void bdrv_aio_cancel_async(BlockAIOCB *acb);
392 
393 /* sg packet commands */
394 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
395 
396 /* Invalidate any cached metadata used by image formats */
397 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
398 void bdrv_invalidate_cache_all(Error **errp);
399 int bdrv_inactivate_all(void);
400 
401 /* Ensure contents are flushed to disk.  */
402 int bdrv_flush(BlockDriverState *bs);
403 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
404 int bdrv_flush_all(void);
405 void bdrv_close_all(void);
406 void bdrv_drain(BlockDriverState *bs);
407 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
408 void bdrv_drain_all_begin(void);
409 void bdrv_drain_all_end(void);
410 void bdrv_drain_all(void);
411 
412 /* Returns NULL when bs == NULL */
413 AioWait *bdrv_get_aio_wait(BlockDriverState *bs);
414 
415 #define BDRV_POLL_WHILE(bs, cond) ({                       \
416     BlockDriverState *bs_ = (bs);                          \
417     AIO_WAIT_WHILE(bdrv_get_aio_wait(bs_),                 \
418                    bdrv_get_aio_context(bs_),              \
419                    cond); })
420 
421 int bdrv_pdiscard(BdrvChild *child, int64_t offset, int bytes);
422 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int bytes);
423 int bdrv_has_zero_init_1(BlockDriverState *bs);
424 int bdrv_has_zero_init(BlockDriverState *bs);
425 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
426 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
427 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
428                       int64_t bytes, int64_t *pnum, int64_t *map,
429                       BlockDriverState **file);
430 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
431                             int64_t offset, int64_t bytes, int64_t *pnum,
432                             int64_t *map, BlockDriverState **file);
433 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
434                       int64_t *pnum);
435 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
436                             int64_t offset, int64_t bytes, int64_t *pnum);
437 
438 bool bdrv_is_read_only(BlockDriverState *bs);
439 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
440                            bool ignore_allow_rdw, Error **errp);
441 int bdrv_set_read_only(BlockDriverState *bs, bool read_only, Error **errp);
442 bool bdrv_is_writable(BlockDriverState *bs);
443 bool bdrv_is_sg(BlockDriverState *bs);
444 bool bdrv_is_inserted(BlockDriverState *bs);
445 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
446 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
447 const char *bdrv_get_format_name(BlockDriverState *bs);
448 BlockDriverState *bdrv_find_node(const char *node_name);
449 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
450 BlockDriverState *bdrv_lookup_bs(const char *device,
451                                  const char *node_name,
452                                  Error **errp);
453 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
454 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
455 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
456 
457 typedef struct BdrvNextIterator {
458     enum {
459         BDRV_NEXT_BACKEND_ROOTS,
460         BDRV_NEXT_MONITOR_OWNED,
461     } phase;
462     BlockBackend *blk;
463     BlockDriverState *bs;
464 } BdrvNextIterator;
465 
466 BlockDriverState *bdrv_first(BdrvNextIterator *it);
467 BlockDriverState *bdrv_next(BdrvNextIterator *it);
468 void bdrv_next_cleanup(BdrvNextIterator *it);
469 
470 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
471 bool bdrv_is_encrypted(BlockDriverState *bs);
472 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
473                          void *opaque);
474 const char *bdrv_get_node_name(const BlockDriverState *bs);
475 const char *bdrv_get_device_name(const BlockDriverState *bs);
476 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
477 int bdrv_get_flags(BlockDriverState *bs);
478 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
479 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs);
480 void bdrv_round_to_clusters(BlockDriverState *bs,
481                             int64_t offset, int64_t bytes,
482                             int64_t *cluster_offset,
483                             int64_t *cluster_bytes);
484 
485 const char *bdrv_get_encrypted_filename(BlockDriverState *bs);
486 void bdrv_get_backing_filename(BlockDriverState *bs,
487                                char *filename, int filename_size);
488 void bdrv_get_full_backing_filename(BlockDriverState *bs,
489                                     char *dest, size_t sz, Error **errp);
490 void bdrv_get_full_backing_filename_from_filename(const char *backed,
491                                                   const char *backing,
492                                                   char *dest, size_t sz,
493                                                   Error **errp);
494 
495 int path_has_protocol(const char *path);
496 int path_is_absolute(const char *path);
497 void path_combine(char *dest, int dest_size,
498                   const char *base_path,
499                   const char *filename);
500 
501 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
502 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
503 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
504                       int64_t pos, int size);
505 
506 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
507                       int64_t pos, int size);
508 
509 void bdrv_img_create(const char *filename, const char *fmt,
510                      const char *base_filename, const char *base_fmt,
511                      char *options, uint64_t img_size, int flags,
512                      bool quiet, Error **errp);
513 
514 /* Returns the alignment in bytes that is required so that no bounce buffer
515  * is required throughout the stack */
516 size_t bdrv_min_mem_align(BlockDriverState *bs);
517 /* Returns optimal alignment in bytes for bounce buffer */
518 size_t bdrv_opt_mem_align(BlockDriverState *bs);
519 void *qemu_blockalign(BlockDriverState *bs, size_t size);
520 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
521 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
522 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
523 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
524 
525 void bdrv_enable_copy_on_read(BlockDriverState *bs);
526 void bdrv_disable_copy_on_read(BlockDriverState *bs);
527 
528 void bdrv_ref(BlockDriverState *bs);
529 void bdrv_unref(BlockDriverState *bs);
530 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
531 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
532                              BlockDriverState *child_bs,
533                              const char *child_name,
534                              const BdrvChildRole *child_role,
535                              Error **errp);
536 
537 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
538 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
539 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
540 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
541 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
542 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
543 
544 #define BLKDBG_EVENT(child, evt) \
545     do { \
546         if (child) { \
547             bdrv_debug_event(child->bs, evt); \
548         } \
549     } while (0)
550 
551 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
552 
553 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
554                            const char *tag);
555 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
556 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
557 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
558 
559 /**
560  * bdrv_get_aio_context:
561  *
562  * Returns: the currently bound #AioContext
563  */
564 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
565 
566 /**
567  * Transfer control to @co in the aio context of @bs
568  */
569 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
570 
571 /**
572  * bdrv_set_aio_context:
573  *
574  * Changes the #AioContext used for fd handlers, timers, and BHs by this
575  * BlockDriverState and all its children.
576  *
577  * This function must be called with iothread lock held.
578  */
579 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
580 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
581 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
582 
583 void bdrv_io_plug(BlockDriverState *bs);
584 void bdrv_io_unplug(BlockDriverState *bs);
585 
586 /**
587  * bdrv_parent_drained_begin:
588  *
589  * Begin a quiesced section of all users of @bs. This is part of
590  * bdrv_drained_begin.
591  */
592 void bdrv_parent_drained_begin(BlockDriverState *bs, BdrvChild *ignore,
593                                bool ignore_bds_parents);
594 
595 /**
596  * bdrv_parent_drained_begin_single:
597  *
598  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
599  * any pending activity to cease.
600  */
601 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
602 
603 /**
604  * bdrv_parent_drained_end:
605  *
606  * End a quiesced section of all users of @bs. This is part of
607  * bdrv_drained_end.
608  */
609 void bdrv_parent_drained_end(BlockDriverState *bs, BdrvChild *ignore,
610                              bool ignore_bds_parents);
611 
612 /**
613  * bdrv_drain_poll:
614  *
615  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
616  * and if @recursive is true its children as well (used for subtree drain).
617  *
618  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
619  * ignore the drain request because they will be drained separately (used for
620  * drain_all).
621  *
622  * This is part of bdrv_drained_begin.
623  */
624 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
625                      BdrvChild *ignore_parent, bool ignore_bds_parents);
626 
627 /**
628  * bdrv_drained_begin:
629  *
630  * Begin a quiesced section for exclusive access to the BDS, by disabling
631  * external request sources including NBD server and device model. Note that
632  * this doesn't block timers or coroutines from submitting more requests, which
633  * means block_job_pause is still necessary.
634  *
635  * This function can be recursive.
636  */
637 void bdrv_drained_begin(BlockDriverState *bs);
638 
639 /**
640  * bdrv_do_drained_begin_quiesce:
641  *
642  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
643  * running requests to complete.
644  */
645 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
646                                    BdrvChild *parent, bool ignore_bds_parents);
647 
648 /**
649  * Like bdrv_drained_begin, but recursively begins a quiesced section for
650  * exclusive access to all child nodes as well.
651  */
652 void bdrv_subtree_drained_begin(BlockDriverState *bs);
653 
654 /**
655  * bdrv_drained_end:
656  *
657  * End a quiescent section started by bdrv_drained_begin().
658  */
659 void bdrv_drained_end(BlockDriverState *bs);
660 
661 /**
662  * End a quiescent section started by bdrv_subtree_drained_begin().
663  */
664 void bdrv_subtree_drained_end(BlockDriverState *bs);
665 
666 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
667                     Error **errp);
668 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
669 
670 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
671                                      uint32_t granularity, Error **errp);
672 /**
673  *
674  * bdrv_register_buf/bdrv_unregister_buf:
675  *
676  * Register/unregister a buffer for I/O. For example, VFIO drivers are
677  * interested to know the memory areas that would later be used for I/O, so
678  * that they can prepare IOMMU mapping etc., to get better performance.
679  */
680 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
681 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
682 
683 /**
684  *
685  * bdrv_co_copy_range:
686  *
687  * Do offloaded copy between two children. If the operation is not implemented
688  * by the driver, or if the backend storage doesn't support it, a negative
689  * error code will be returned.
690  *
691  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
692  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
693  * calling copy_file_range(2)) after the first error, thus it should fall back
694  * to a read+write path in the caller level.
695  *
696  * @src: Source child to copy data from
697  * @src_offset: offset in @src image to read data
698  * @dst: Destination child to copy data to
699  * @dst_offset: offset in @dst image to write data
700  * @bytes: number of bytes to copy
701  * @flags: request flags. Supported flags:
702  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
703  *                               write on @dst as if bdrv_co_pwrite_zeroes is
704  *                               called. Used to simplify caller code, or
705  *                               during BlockDriver.bdrv_co_copy_range_from()
706  *                               recursion.
707  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
708  *                                   requests currently in flight.
709  *
710  * Returns: 0 if succeeded; negative error code if failed.
711  **/
712 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
713                                     BdrvChild *dst, uint64_t dst_offset,
714                                     uint64_t bytes, BdrvRequestFlags read_flags,
715                                     BdrvRequestFlags write_flags);
716 #endif
717