xref: /openbmc/qemu/include/block/block.h (revision 795c40b8)
1 #ifndef BLOCK_H
2 #define BLOCK_H
3 
4 #include "block/aio.h"
5 #include "qemu/iov.h"
6 #include "qemu/option.h"
7 #include "qemu/coroutine.h"
8 #include "block/accounting.h"
9 #include "block/dirty-bitmap.h"
10 #include "block/blockjob.h"
11 #include "qapi/qmp/qobject.h"
12 #include "qapi-types.h"
13 #include "qemu/hbitmap.h"
14 
15 /* block.c */
16 typedef struct BlockDriver BlockDriver;
17 typedef struct BdrvChild BdrvChild;
18 typedef struct BdrvChildRole BdrvChildRole;
19 
20 typedef struct BlockDriverInfo {
21     /* in bytes, 0 if irrelevant */
22     int cluster_size;
23     /* offset at which the VM state can be saved (0 if not possible) */
24     int64_t vm_state_offset;
25     bool is_dirty;
26     /*
27      * True if unallocated blocks read back as zeroes. This is equivalent
28      * to the LBPRZ flag in the SCSI logical block provisioning page.
29      */
30     bool unallocated_blocks_are_zero;
31     /*
32      * True if the driver can optimize writing zeroes by unmapping
33      * sectors. This is equivalent to the BLKDISCARDZEROES ioctl in Linux
34      * with the difference that in qemu a discard is allowed to silently
35      * fail. Therefore we have to use bdrv_pwrite_zeroes with the
36      * BDRV_REQ_MAY_UNMAP flag for an optimized zero write with unmapping.
37      * After this call the driver has to guarantee that the contents read
38      * back as zero. It is additionally required that the block device is
39      * opened with BDRV_O_UNMAP flag for this to work.
40      */
41     bool can_write_zeroes_with_unmap;
42     /*
43      * True if this block driver only supports compressed writes
44      */
45     bool needs_compressed_writes;
46 } BlockDriverInfo;
47 
48 typedef struct BlockFragInfo {
49     uint64_t allocated_clusters;
50     uint64_t total_clusters;
51     uint64_t fragmented_clusters;
52     uint64_t compressed_clusters;
53 } BlockFragInfo;
54 
55 typedef enum {
56     BDRV_REQ_COPY_ON_READ       = 0x1,
57     BDRV_REQ_ZERO_WRITE         = 0x2,
58     /* The BDRV_REQ_MAY_UNMAP flag is used to indicate that the block driver
59      * is allowed to optimize a write zeroes request by unmapping (discarding)
60      * blocks if it is guaranteed that the result will read back as
61      * zeroes. The flag is only passed to the driver if the block device is
62      * opened with BDRV_O_UNMAP.
63      */
64     BDRV_REQ_MAY_UNMAP          = 0x4,
65     BDRV_REQ_NO_SERIALISING     = 0x8,
66     BDRV_REQ_FUA                = 0x10,
67     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
68 
69     /* Mask of valid flags */
70     BDRV_REQ_MASK               = 0x3f,
71 } BdrvRequestFlags;
72 
73 typedef struct BlockSizes {
74     uint32_t phys;
75     uint32_t log;
76 } BlockSizes;
77 
78 typedef struct HDGeometry {
79     uint32_t heads;
80     uint32_t sectors;
81     uint32_t cylinders;
82 } HDGeometry;
83 
84 #define BDRV_O_RDWR        0x0002
85 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
86 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
87 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
88 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
89 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
90 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
91 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
92 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
93 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
94 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
95 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
96 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
97 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
98                                       select an appropriate protocol driver,
99                                       ignoring the format layer */
100 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
101 
102 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
103 
104 
105 /* Option names of options parsed by the block layer */
106 
107 #define BDRV_OPT_CACHE_WB       "cache.writeback"
108 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
109 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
110 #define BDRV_OPT_READ_ONLY      "read-only"
111 #define BDRV_OPT_DISCARD        "discard"
112 #define BDRV_OPT_FORCE_SHARE    "force-share"
113 
114 
115 #define BDRV_SECTOR_BITS   9
116 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
117 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
118 
119 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
120                                      INT_MAX >> BDRV_SECTOR_BITS)
121 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
122 
123 /*
124  * Allocation status flags for bdrv_get_block_status() and friends.
125  *
126  * Public flags:
127  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
128  * BDRV_BLOCK_ZERO: offset reads as zero
129  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
130  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
131  *                       layer (short for DATA || ZERO), set by block layer
132  *
133  * Internal flag:
134  * BDRV_BLOCK_RAW: used internally to indicate that the request was
135  *                 answered by a passthrough driver such as raw and that the
136  *                 block layer should recompute the answer from bs->file.
137  *
138  * If BDRV_BLOCK_OFFSET_VALID is set, bits 9-62 (BDRV_BLOCK_OFFSET_MASK)
139  * represent the offset in the returned BDS that is allocated for the
140  * corresponding raw data; however, whether that offset actually contains
141  * data also depends on BDRV_BLOCK_DATA and BDRV_BLOCK_ZERO, as follows:
142  *
143  * DATA ZERO OFFSET_VALID
144  *  t    t        t       sectors read as zero, returned file is zero at offset
145  *  t    f        t       sectors read as valid from file at offset
146  *  f    t        t       sectors preallocated, read as zero, returned file not
147  *                        necessarily zero at offset
148  *  f    f        t       sectors preallocated but read from backing_hd,
149  *                        returned file contains garbage at offset
150  *  t    t        f       sectors preallocated, read as zero, unknown offset
151  *  t    f        f       sectors read from unknown file or offset
152  *  f    t        f       not allocated or unknown offset, read as zero
153  *  f    f        f       not allocated or unknown offset, read from backing_hd
154  */
155 #define BDRV_BLOCK_DATA         0x01
156 #define BDRV_BLOCK_ZERO         0x02
157 #define BDRV_BLOCK_OFFSET_VALID 0x04
158 #define BDRV_BLOCK_RAW          0x08
159 #define BDRV_BLOCK_ALLOCATED    0x10
160 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
161 
162 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
163 
164 typedef struct BDRVReopenState {
165     BlockDriverState *bs;
166     int flags;
167     QDict *options;
168     QDict *explicit_options;
169     void *opaque;
170 } BDRVReopenState;
171 
172 /*
173  * Block operation types
174  */
175 typedef enum BlockOpType {
176     BLOCK_OP_TYPE_BACKUP_SOURCE,
177     BLOCK_OP_TYPE_BACKUP_TARGET,
178     BLOCK_OP_TYPE_CHANGE,
179     BLOCK_OP_TYPE_COMMIT_SOURCE,
180     BLOCK_OP_TYPE_COMMIT_TARGET,
181     BLOCK_OP_TYPE_DATAPLANE,
182     BLOCK_OP_TYPE_DRIVE_DEL,
183     BLOCK_OP_TYPE_EJECT,
184     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
185     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
186     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
187     BLOCK_OP_TYPE_MIRROR_SOURCE,
188     BLOCK_OP_TYPE_MIRROR_TARGET,
189     BLOCK_OP_TYPE_RESIZE,
190     BLOCK_OP_TYPE_STREAM,
191     BLOCK_OP_TYPE_REPLACE,
192     BLOCK_OP_TYPE_MAX,
193 } BlockOpType;
194 
195 /* Block node permission constants */
196 enum {
197     /**
198      * A user that has the "permission" of consistent reads is guaranteed that
199      * their view of the contents of the block device is complete and
200      * self-consistent, representing the contents of a disk at a specific
201      * point.
202      *
203      * For most block devices (including their backing files) this is true, but
204      * the property cannot be maintained in a few situations like for
205      * intermediate nodes of a commit block job.
206      */
207     BLK_PERM_CONSISTENT_READ    = 0x01,
208 
209     /** This permission is required to change the visible disk contents. */
210     BLK_PERM_WRITE              = 0x02,
211 
212     /**
213      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
214      * required for writes to the block node when the caller promises that
215      * the visible disk content doesn't change.
216      */
217     BLK_PERM_WRITE_UNCHANGED    = 0x04,
218 
219     /** This permission is required to change the size of a block node. */
220     BLK_PERM_RESIZE             = 0x08,
221 
222     /**
223      * This permission is required to change the node that this BdrvChild
224      * points to.
225      */
226     BLK_PERM_GRAPH_MOD          = 0x10,
227 
228     BLK_PERM_ALL                = 0x1f,
229 };
230 
231 char *bdrv_perm_names(uint64_t perm);
232 
233 /* disk I/O throttling */
234 void bdrv_init(void);
235 void bdrv_init_with_whitelist(void);
236 bool bdrv_uses_whitelist(void);
237 BlockDriver *bdrv_find_protocol(const char *filename,
238                                 bool allow_protocol_prefix,
239                                 Error **errp);
240 BlockDriver *bdrv_find_format(const char *format_name);
241 int bdrv_create(BlockDriver *drv, const char* filename,
242                 QemuOpts *opts, Error **errp);
243 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
244 BlockDriverState *bdrv_new(void);
245 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
246                  Error **errp);
247 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
248                        Error **errp);
249 
250 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
251 int bdrv_parse_discard_flags(const char *mode, int *flags);
252 BdrvChild *bdrv_open_child(const char *filename,
253                            QDict *options, const char *bdref_key,
254                            BlockDriverState* parent,
255                            const BdrvChildRole *child_role,
256                            bool allow_none, Error **errp);
257 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
258                          Error **errp);
259 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
260                            const char *bdref_key, Error **errp);
261 BlockDriverState *bdrv_open(const char *filename, const char *reference,
262                             QDict *options, int flags, Error **errp);
263 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
264                                        int flags, Error **errp);
265 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
266                                     BlockDriverState *bs,
267                                     QDict *options, int flags);
268 int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp);
269 int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp);
270 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
271                         BlockReopenQueue *queue, Error **errp);
272 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
273 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
274 int bdrv_read(BdrvChild *child, int64_t sector_num,
275               uint8_t *buf, int nb_sectors);
276 int bdrv_write(BdrvChild *child, int64_t sector_num,
277                const uint8_t *buf, int nb_sectors);
278 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
279                        int count, BdrvRequestFlags flags);
280 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
281 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
282 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
283 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
284 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
285 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
286                      const void *buf, int count);
287 int coroutine_fn bdrv_co_readv(BdrvChild *child, int64_t sector_num,
288                                int nb_sectors, QEMUIOVector *qiov);
289 int coroutine_fn bdrv_co_writev(BdrvChild *child, int64_t sector_num,
290                                int nb_sectors, QEMUIOVector *qiov);
291 /*
292  * Efficiently zero a region of the disk image.  Note that this is a regular
293  * I/O request like read or write and should have a reasonable size.  This
294  * function is not suitable for zeroing the entire image in a single request
295  * because it may allocate memory for the entire region.
296  */
297 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
298                                        int count, BdrvRequestFlags flags);
299 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
300     const char *backing_file);
301 int bdrv_get_backing_file_depth(BlockDriverState *bs);
302 void bdrv_refresh_filename(BlockDriverState *bs);
303 int bdrv_truncate(BdrvChild *child, int64_t offset, Error **errp);
304 int64_t bdrv_nb_sectors(BlockDriverState *bs);
305 int64_t bdrv_getlength(BlockDriverState *bs);
306 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
307 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
308 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
309 int bdrv_commit(BlockDriverState *bs);
310 int bdrv_change_backing_file(BlockDriverState *bs,
311     const char *backing_file, const char *backing_fmt);
312 void bdrv_register(BlockDriver *bdrv);
313 int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top,
314                            BlockDriverState *base,
315                            const char *backing_file_str);
316 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
317                                     BlockDriverState *bs);
318 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
319 
320 
321 typedef struct BdrvCheckResult {
322     int corruptions;
323     int leaks;
324     int check_errors;
325     int corruptions_fixed;
326     int leaks_fixed;
327     int64_t image_end_offset;
328     BlockFragInfo bfi;
329 } BdrvCheckResult;
330 
331 typedef enum {
332     BDRV_FIX_LEAKS    = 1,
333     BDRV_FIX_ERRORS   = 2,
334 } BdrvCheckMode;
335 
336 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
337 
338 /* The units of offset and total_work_size may be chosen arbitrarily by the
339  * block driver; total_work_size may change during the course of the amendment
340  * operation */
341 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
342                                       int64_t total_work_size, void *opaque);
343 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
344                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque);
345 
346 /* external snapshots */
347 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
348                                       BlockDriverState *candidate);
349 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
350 
351 /* check if a named node can be replaced when doing drive-mirror */
352 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
353                                         const char *node_name, Error **errp);
354 
355 /* async block I/O */
356 BlockAIOCB *bdrv_aio_readv(BdrvChild *child, int64_t sector_num,
357                            QEMUIOVector *iov, int nb_sectors,
358                            BlockCompletionFunc *cb, void *opaque);
359 BlockAIOCB *bdrv_aio_writev(BdrvChild *child, int64_t sector_num,
360                             QEMUIOVector *iov, int nb_sectors,
361                             BlockCompletionFunc *cb, void *opaque);
362 BlockAIOCB *bdrv_aio_flush(BlockDriverState *bs,
363                            BlockCompletionFunc *cb, void *opaque);
364 void bdrv_aio_cancel(BlockAIOCB *acb);
365 void bdrv_aio_cancel_async(BlockAIOCB *acb);
366 
367 /* sg packet commands */
368 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
369 
370 /* Invalidate any cached metadata used by image formats */
371 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
372 void bdrv_invalidate_cache_all(Error **errp);
373 int bdrv_inactivate_all(void);
374 
375 /* Ensure contents are flushed to disk.  */
376 int bdrv_flush(BlockDriverState *bs);
377 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
378 int bdrv_flush_all(void);
379 void bdrv_close_all(void);
380 void bdrv_drain(BlockDriverState *bs);
381 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
382 void bdrv_drain_all_begin(void);
383 void bdrv_drain_all_end(void);
384 void bdrv_drain_all(void);
385 
386 #define BDRV_POLL_WHILE(bs, cond) ({                       \
387     bool waited_ = false;                                  \
388     bool busy_ = true;                                     \
389     BlockDriverState *bs_ = (bs);                          \
390     AioContext *ctx_ = bdrv_get_aio_context(bs_);          \
391     if (aio_context_in_iothread(ctx_)) {                   \
392         while ((cond) || busy_) {                          \
393             busy_ = aio_poll(ctx_, (cond));                \
394             waited_ |= !!(cond) | busy_;                   \
395         }                                                  \
396     } else {                                               \
397         assert(qemu_get_current_aio_context() ==           \
398                qemu_get_aio_context());                    \
399         /* Ask bdrv_dec_in_flight to wake up the main      \
400          * QEMU AioContext.  Extra I/O threads never take  \
401          * other I/O threads' AioContexts (see for example \
402          * block_job_defer_to_main_loop for how to do it). \
403          */                                                \
404         assert(!bs_->wakeup);                              \
405         bs_->wakeup = true;                                \
406         while (busy_) {                                    \
407             if ((cond)) {                                  \
408                 waited_ = busy_ = true;                    \
409                 aio_context_release(ctx_);                 \
410                 aio_poll(qemu_get_aio_context(), true);    \
411                 aio_context_acquire(ctx_);                 \
412             } else {                                       \
413                 busy_ = aio_poll(ctx_, false);             \
414                 waited_ |= busy_;                          \
415             }                                              \
416         }                                                  \
417         bs_->wakeup = false;                               \
418     }                                                      \
419     waited_; })
420 
421 int bdrv_pdiscard(BlockDriverState *bs, int64_t offset, int count);
422 int bdrv_co_pdiscard(BlockDriverState *bs, int64_t offset, int count);
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 int64_t bdrv_get_block_status(BlockDriverState *bs, int64_t sector_num,
428                               int nb_sectors, int *pnum,
429                               BlockDriverState **file);
430 int64_t bdrv_get_block_status_above(BlockDriverState *bs,
431                                     BlockDriverState *base,
432                                     int64_t sector_num,
433                                     int nb_sectors, int *pnum,
434                                     BlockDriverState **file);
435 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
436                       int *pnum);
437 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
438                             int64_t sector_num, int nb_sectors, int *pnum);
439 
440 bool bdrv_is_read_only(BlockDriverState *bs);
441 bool bdrv_is_writable(BlockDriverState *bs);
442 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only, Error **errp);
443 int bdrv_set_read_only(BlockDriverState *bs, bool read_only, Error **errp);
444 bool bdrv_is_sg(BlockDriverState *bs);
445 bool bdrv_is_inserted(BlockDriverState *bs);
446 int bdrv_media_changed(BlockDriverState *bs);
447 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
448 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
449 const char *bdrv_get_format_name(BlockDriverState *bs);
450 BlockDriverState *bdrv_find_node(const char *node_name);
451 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
452 BlockDriverState *bdrv_lookup_bs(const char *device,
453                                  const char *node_name,
454                                  Error **errp);
455 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
456 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
457 
458 typedef struct BdrvNextIterator {
459     enum {
460         BDRV_NEXT_BACKEND_ROOTS,
461         BDRV_NEXT_MONITOR_OWNED,
462     } phase;
463     BlockBackend *blk;
464     BlockDriverState *bs;
465 } BdrvNextIterator;
466 
467 BlockDriverState *bdrv_first(BdrvNextIterator *it);
468 BlockDriverState *bdrv_next(BdrvNextIterator *it);
469 
470 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
471 bool bdrv_is_encrypted(BlockDriverState *bs);
472 bool bdrv_key_required(BlockDriverState *bs);
473 int bdrv_set_key(BlockDriverState *bs, const char *key);
474 void bdrv_add_key(BlockDriverState *bs, const char *key, Error **errp);
475 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
476                          void *opaque);
477 const char *bdrv_get_node_name(const BlockDriverState *bs);
478 const char *bdrv_get_device_name(const BlockDriverState *bs);
479 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
480 int bdrv_get_flags(BlockDriverState *bs);
481 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
482 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs);
483 void bdrv_round_sectors_to_clusters(BlockDriverState *bs,
484                                     int64_t sector_num, int nb_sectors,
485                                     int64_t *cluster_sector_num,
486                                     int *cluster_nb_sectors);
487 void bdrv_round_to_clusters(BlockDriverState *bs,
488                             int64_t offset, unsigned int bytes,
489                             int64_t *cluster_offset,
490                             unsigned int *cluster_bytes);
491 
492 const char *bdrv_get_encrypted_filename(BlockDriverState *bs);
493 void bdrv_get_backing_filename(BlockDriverState *bs,
494                                char *filename, int filename_size);
495 void bdrv_get_full_backing_filename(BlockDriverState *bs,
496                                     char *dest, size_t sz, Error **errp);
497 void bdrv_get_full_backing_filename_from_filename(const char *backed,
498                                                   const char *backing,
499                                                   char *dest, size_t sz,
500                                                   Error **errp);
501 
502 int path_has_protocol(const char *path);
503 int path_is_absolute(const char *path);
504 void path_combine(char *dest, int dest_size,
505                   const char *base_path,
506                   const char *filename);
507 
508 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
509 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
510 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
511                       int64_t pos, int size);
512 
513 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
514                       int64_t pos, int size);
515 
516 void bdrv_img_create(const char *filename, const char *fmt,
517                      const char *base_filename, const char *base_fmt,
518                      char *options, uint64_t img_size, int flags,
519                      bool quiet, Error **errp);
520 
521 /* Returns the alignment in bytes that is required so that no bounce buffer
522  * is required throughout the stack */
523 size_t bdrv_min_mem_align(BlockDriverState *bs);
524 /* Returns optimal alignment in bytes for bounce buffer */
525 size_t bdrv_opt_mem_align(BlockDriverState *bs);
526 void *qemu_blockalign(BlockDriverState *bs, size_t size);
527 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
528 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
529 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
530 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
531 
532 void bdrv_enable_copy_on_read(BlockDriverState *bs);
533 void bdrv_disable_copy_on_read(BlockDriverState *bs);
534 
535 void bdrv_ref(BlockDriverState *bs);
536 void bdrv_unref(BlockDriverState *bs);
537 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
538 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
539                              BlockDriverState *child_bs,
540                              const char *child_name,
541                              const BdrvChildRole *child_role,
542                              Error **errp);
543 
544 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
545 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
546 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
547 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
548 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
549 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
550 
551 #define BLKDBG_EVENT(child, evt) \
552     do { \
553         if (child) { \
554             bdrv_debug_event(child->bs, evt); \
555         } \
556     } while (0)
557 
558 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
559 
560 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
561                            const char *tag);
562 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
563 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
564 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
565 
566 /**
567  * bdrv_get_aio_context:
568  *
569  * Returns: the currently bound #AioContext
570  */
571 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
572 
573 /**
574  * Transfer control to @co in the aio context of @bs
575  */
576 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
577 
578 /**
579  * bdrv_set_aio_context:
580  *
581  * Changes the #AioContext used for fd handlers, timers, and BHs by this
582  * BlockDriverState and all its children.
583  *
584  * This function must be called with iothread lock held.
585  */
586 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
587 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
588 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
589 
590 void bdrv_io_plug(BlockDriverState *bs);
591 void bdrv_io_unplug(BlockDriverState *bs);
592 
593 /**
594  * bdrv_parent_drained_begin:
595  *
596  * Begin a quiesced section of all users of @bs. This is part of
597  * bdrv_drained_begin.
598  */
599 void bdrv_parent_drained_begin(BlockDriverState *bs);
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);
608 
609 /**
610  * bdrv_drained_begin:
611  *
612  * Begin a quiesced section for exclusive access to the BDS, by disabling
613  * external request sources including NBD server and device model. Note that
614  * this doesn't block timers or coroutines from submitting more requests, which
615  * means block_job_pause is still necessary.
616  *
617  * This function can be recursive.
618  */
619 void bdrv_drained_begin(BlockDriverState *bs);
620 
621 /**
622  * bdrv_drained_end:
623  *
624  * End a quiescent section started by bdrv_drained_begin().
625  */
626 void bdrv_drained_end(BlockDriverState *bs);
627 
628 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
629                     Error **errp);
630 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
631 
632 #endif
633