xref: /openbmc/qemu/include/block/block.h (revision 8f0a3716)
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_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  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this layer
133  *
134  * Internal flag:
135  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
136  *                 that the block layer recompute the answer from the returned
137  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
138  *
139  * If BDRV_BLOCK_OFFSET_VALID is set, bits 9-62 (BDRV_BLOCK_OFFSET_MASK) of
140  * the return value (old interface) or the entire map parameter (new
141  * interface) represent the offset in the returned BDS that is allocated for
142  * the corresponding raw data.  However, whether that offset actually
143  * contains data also depends on BDRV_BLOCK_DATA, as follows:
144  *
145  * DATA ZERO OFFSET_VALID
146  *  t    t        t       sectors read as zero, returned file is zero at offset
147  *  t    f        t       sectors read as valid from file at offset
148  *  f    t        t       sectors preallocated, read as zero, returned file not
149  *                        necessarily zero at offset
150  *  f    f        t       sectors preallocated but read from backing_hd,
151  *                        returned file contains garbage at offset
152  *  t    t        f       sectors preallocated, read as zero, unknown offset
153  *  t    f        f       sectors read from unknown file or offset
154  *  f    t        f       not allocated or unknown offset, read as zero
155  *  f    f        f       not allocated or unknown offset, read from backing_hd
156  */
157 #define BDRV_BLOCK_DATA         0x01
158 #define BDRV_BLOCK_ZERO         0x02
159 #define BDRV_BLOCK_OFFSET_VALID 0x04
160 #define BDRV_BLOCK_RAW          0x08
161 #define BDRV_BLOCK_ALLOCATED    0x10
162 #define BDRV_BLOCK_EOF          0x20
163 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
164 
165 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
166 
167 typedef struct BDRVReopenState {
168     BlockDriverState *bs;
169     int flags;
170     uint64_t perm, shared_perm;
171     QDict *options;
172     QDict *explicit_options;
173     void *opaque;
174 } BDRVReopenState;
175 
176 /*
177  * Block operation types
178  */
179 typedef enum BlockOpType {
180     BLOCK_OP_TYPE_BACKUP_SOURCE,
181     BLOCK_OP_TYPE_BACKUP_TARGET,
182     BLOCK_OP_TYPE_CHANGE,
183     BLOCK_OP_TYPE_COMMIT_SOURCE,
184     BLOCK_OP_TYPE_COMMIT_TARGET,
185     BLOCK_OP_TYPE_DATAPLANE,
186     BLOCK_OP_TYPE_DRIVE_DEL,
187     BLOCK_OP_TYPE_EJECT,
188     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
189     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
190     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
191     BLOCK_OP_TYPE_MIRROR_SOURCE,
192     BLOCK_OP_TYPE_MIRROR_TARGET,
193     BLOCK_OP_TYPE_RESIZE,
194     BLOCK_OP_TYPE_STREAM,
195     BLOCK_OP_TYPE_REPLACE,
196     BLOCK_OP_TYPE_MAX,
197 } BlockOpType;
198 
199 /* Block node permission constants */
200 enum {
201     /**
202      * A user that has the "permission" of consistent reads is guaranteed that
203      * their view of the contents of the block device is complete and
204      * self-consistent, representing the contents of a disk at a specific
205      * point.
206      *
207      * For most block devices (including their backing files) this is true, but
208      * the property cannot be maintained in a few situations like for
209      * intermediate nodes of a commit block job.
210      */
211     BLK_PERM_CONSISTENT_READ    = 0x01,
212 
213     /** This permission is required to change the visible disk contents. */
214     BLK_PERM_WRITE              = 0x02,
215 
216     /**
217      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
218      * required for writes to the block node when the caller promises that
219      * the visible disk content doesn't change.
220      */
221     BLK_PERM_WRITE_UNCHANGED    = 0x04,
222 
223     /** This permission is required to change the size of a block node. */
224     BLK_PERM_RESIZE             = 0x08,
225 
226     /**
227      * This permission is required to change the node that this BdrvChild
228      * points to.
229      */
230     BLK_PERM_GRAPH_MOD          = 0x10,
231 
232     BLK_PERM_ALL                = 0x1f,
233 };
234 
235 char *bdrv_perm_names(uint64_t perm);
236 
237 /* disk I/O throttling */
238 void bdrv_init(void);
239 void bdrv_init_with_whitelist(void);
240 bool bdrv_uses_whitelist(void);
241 BlockDriver *bdrv_find_protocol(const char *filename,
242                                 bool allow_protocol_prefix,
243                                 Error **errp);
244 BlockDriver *bdrv_find_format(const char *format_name);
245 int bdrv_create(BlockDriver *drv, const char* filename,
246                 QemuOpts *opts, Error **errp);
247 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
248 BlockDriverState *bdrv_new(void);
249 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
250                  Error **errp);
251 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
252                        Error **errp);
253 
254 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
255 int bdrv_parse_discard_flags(const char *mode, int *flags);
256 BdrvChild *bdrv_open_child(const char *filename,
257                            QDict *options, const char *bdref_key,
258                            BlockDriverState* parent,
259                            const BdrvChildRole *child_role,
260                            bool allow_none, Error **errp);
261 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
262                          Error **errp);
263 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
264                            const char *bdref_key, Error **errp);
265 BlockDriverState *bdrv_open(const char *filename, const char *reference,
266                             QDict *options, int flags, Error **errp);
267 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
268                                        int flags, Error **errp);
269 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
270                                     BlockDriverState *bs,
271                                     QDict *options, int flags);
272 int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp);
273 int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp);
274 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
275                         BlockReopenQueue *queue, Error **errp);
276 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
277 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
278 int bdrv_read(BdrvChild *child, int64_t sector_num,
279               uint8_t *buf, int nb_sectors);
280 int bdrv_write(BdrvChild *child, int64_t sector_num,
281                const uint8_t *buf, int nb_sectors);
282 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
283                        int bytes, BdrvRequestFlags flags);
284 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
285 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
286 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
287 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
288 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
289 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
290                      const void *buf, int count);
291 int coroutine_fn bdrv_co_readv(BdrvChild *child, int64_t sector_num,
292                                int nb_sectors, QEMUIOVector *qiov);
293 int coroutine_fn bdrv_co_writev(BdrvChild *child, int64_t sector_num,
294                                int nb_sectors, QEMUIOVector *qiov);
295 /*
296  * Efficiently zero a region of the disk image.  Note that this is a regular
297  * I/O request like read or write and should have a reasonable size.  This
298  * function is not suitable for zeroing the entire image in a single request
299  * because it may allocate memory for the entire region.
300  */
301 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
302                                        int bytes, BdrvRequestFlags flags);
303 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
304     const char *backing_file);
305 void bdrv_refresh_filename(BlockDriverState *bs);
306 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
307                   Error **errp);
308 int64_t bdrv_nb_sectors(BlockDriverState *bs);
309 int64_t bdrv_getlength(BlockDriverState *bs);
310 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
311 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
312                                BlockDriverState *in_bs, Error **errp);
313 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
314 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
315 int bdrv_commit(BlockDriverState *bs);
316 int bdrv_change_backing_file(BlockDriverState *bs,
317     const char *backing_file, const char *backing_fmt);
318 void bdrv_register(BlockDriver *bdrv);
319 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
320                            const char *backing_file_str);
321 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
322                                     BlockDriverState *bs);
323 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
324 
325 
326 typedef struct BdrvCheckResult {
327     int corruptions;
328     int leaks;
329     int check_errors;
330     int corruptions_fixed;
331     int leaks_fixed;
332     int64_t image_end_offset;
333     BlockFragInfo bfi;
334 } BdrvCheckResult;
335 
336 typedef enum {
337     BDRV_FIX_LEAKS    = 1,
338     BDRV_FIX_ERRORS   = 2,
339 } BdrvCheckMode;
340 
341 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
342 
343 /* The units of offset and total_work_size may be chosen arbitrarily by the
344  * block driver; total_work_size may change during the course of the amendment
345  * operation */
346 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
347                                       int64_t total_work_size, void *opaque);
348 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
349                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque);
350 
351 /* external snapshots */
352 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
353                                       BlockDriverState *candidate);
354 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
355 
356 /* check if a named node can be replaced when doing drive-mirror */
357 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
358                                         const char *node_name, Error **errp);
359 
360 /* async block I/O */
361 void bdrv_aio_cancel(BlockAIOCB *acb);
362 void bdrv_aio_cancel_async(BlockAIOCB *acb);
363 
364 /* sg packet commands */
365 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
366 
367 /* Invalidate any cached metadata used by image formats */
368 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
369 void bdrv_invalidate_cache_all(Error **errp);
370 int bdrv_inactivate_all(void);
371 
372 /* Ensure contents are flushed to disk.  */
373 int bdrv_flush(BlockDriverState *bs);
374 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
375 int bdrv_flush_all(void);
376 void bdrv_close_all(void);
377 void bdrv_drain(BlockDriverState *bs);
378 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
379 void bdrv_drain_all_begin(void);
380 void bdrv_drain_all_end(void);
381 void bdrv_drain_all(void);
382 
383 #define BDRV_POLL_WHILE(bs, cond) ({                       \
384     bool waited_ = false;                                  \
385     bool busy_ = true;                                     \
386     BlockDriverState *bs_ = (bs);                          \
387     AioContext *ctx_ = bdrv_get_aio_context(bs_);          \
388     if (aio_context_in_iothread(ctx_)) {                   \
389         while ((cond) || busy_) {                          \
390             busy_ = aio_poll(ctx_, (cond));                \
391             waited_ |= !!(cond) | busy_;                   \
392         }                                                  \
393     } else {                                               \
394         assert(qemu_get_current_aio_context() ==           \
395                qemu_get_aio_context());                    \
396         /* Ask bdrv_dec_in_flight to wake up the main      \
397          * QEMU AioContext.  Extra I/O threads never take  \
398          * other I/O threads' AioContexts (see for example \
399          * block_job_defer_to_main_loop for how to do it). \
400          */                                                \
401         assert(!bs_->wakeup);                              \
402         /* Set bs->wakeup before evaluating cond.  */      \
403         atomic_mb_set(&bs_->wakeup, true);                 \
404         while (busy_) {                                    \
405             if ((cond)) {                                  \
406                 waited_ = busy_ = true;                    \
407                 aio_context_release(ctx_);                 \
408                 aio_poll(qemu_get_aio_context(), true);    \
409                 aio_context_acquire(ctx_);                 \
410             } else {                                       \
411                 busy_ = aio_poll(ctx_, false);             \
412                 waited_ |= busy_;                          \
413             }                                              \
414         }                                                  \
415         atomic_set(&bs_->wakeup, false);                   \
416     }                                                      \
417     waited_; })
418 
419 int bdrv_pdiscard(BlockDriverState *bs, int64_t offset, int bytes);
420 int bdrv_co_pdiscard(BlockDriverState *bs, 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_sg(BlockDriverState *bs);
441 bool bdrv_is_inserted(BlockDriverState *bs);
442 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
443 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
444 const char *bdrv_get_format_name(BlockDriverState *bs);
445 BlockDriverState *bdrv_find_node(const char *node_name);
446 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
447 BlockDriverState *bdrv_lookup_bs(const char *device,
448                                  const char *node_name,
449                                  Error **errp);
450 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
451 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
452 
453 typedef struct BdrvNextIterator {
454     enum {
455         BDRV_NEXT_BACKEND_ROOTS,
456         BDRV_NEXT_MONITOR_OWNED,
457     } phase;
458     BlockBackend *blk;
459     BlockDriverState *bs;
460 } BdrvNextIterator;
461 
462 BlockDriverState *bdrv_first(BdrvNextIterator *it);
463 BlockDriverState *bdrv_next(BdrvNextIterator *it);
464 void bdrv_next_cleanup(BdrvNextIterator *it);
465 
466 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
467 bool bdrv_is_encrypted(BlockDriverState *bs);
468 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
469                          void *opaque);
470 const char *bdrv_get_node_name(const BlockDriverState *bs);
471 const char *bdrv_get_device_name(const BlockDriverState *bs);
472 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
473 int bdrv_get_flags(BlockDriverState *bs);
474 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
475 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs);
476 void bdrv_round_to_clusters(BlockDriverState *bs,
477                             int64_t offset, int64_t bytes,
478                             int64_t *cluster_offset,
479                             int64_t *cluster_bytes);
480 
481 const char *bdrv_get_encrypted_filename(BlockDriverState *bs);
482 void bdrv_get_backing_filename(BlockDriverState *bs,
483                                char *filename, int filename_size);
484 void bdrv_get_full_backing_filename(BlockDriverState *bs,
485                                     char *dest, size_t sz, Error **errp);
486 void bdrv_get_full_backing_filename_from_filename(const char *backed,
487                                                   const char *backing,
488                                                   char *dest, size_t sz,
489                                                   Error **errp);
490 
491 int path_has_protocol(const char *path);
492 int path_is_absolute(const char *path);
493 void path_combine(char *dest, int dest_size,
494                   const char *base_path,
495                   const char *filename);
496 
497 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
498 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
499 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
500                       int64_t pos, int size);
501 
502 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
503                       int64_t pos, int size);
504 
505 void bdrv_img_create(const char *filename, const char *fmt,
506                      const char *base_filename, const char *base_fmt,
507                      char *options, uint64_t img_size, int flags,
508                      bool quiet, Error **errp);
509 
510 /* Returns the alignment in bytes that is required so that no bounce buffer
511  * is required throughout the stack */
512 size_t bdrv_min_mem_align(BlockDriverState *bs);
513 /* Returns optimal alignment in bytes for bounce buffer */
514 size_t bdrv_opt_mem_align(BlockDriverState *bs);
515 void *qemu_blockalign(BlockDriverState *bs, size_t size);
516 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
517 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
518 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
519 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
520 
521 void bdrv_enable_copy_on_read(BlockDriverState *bs);
522 void bdrv_disable_copy_on_read(BlockDriverState *bs);
523 
524 void bdrv_ref(BlockDriverState *bs);
525 void bdrv_unref(BlockDriverState *bs);
526 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
527 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
528                              BlockDriverState *child_bs,
529                              const char *child_name,
530                              const BdrvChildRole *child_role,
531                              Error **errp);
532 
533 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
534 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
535 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
536 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
537 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
538 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
539 
540 #define BLKDBG_EVENT(child, evt) \
541     do { \
542         if (child) { \
543             bdrv_debug_event(child->bs, evt); \
544         } \
545     } while (0)
546 
547 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
548 
549 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
550                            const char *tag);
551 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
552 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
553 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
554 
555 /**
556  * bdrv_get_aio_context:
557  *
558  * Returns: the currently bound #AioContext
559  */
560 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
561 
562 /**
563  * Transfer control to @co in the aio context of @bs
564  */
565 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
566 
567 /**
568  * bdrv_set_aio_context:
569  *
570  * Changes the #AioContext used for fd handlers, timers, and BHs by this
571  * BlockDriverState and all its children.
572  *
573  * This function must be called with iothread lock held.
574  */
575 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
576 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
577 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
578 
579 void bdrv_io_plug(BlockDriverState *bs);
580 void bdrv_io_unplug(BlockDriverState *bs);
581 
582 /**
583  * bdrv_parent_drained_begin:
584  *
585  * Begin a quiesced section of all users of @bs. This is part of
586  * bdrv_drained_begin.
587  */
588 void bdrv_parent_drained_begin(BlockDriverState *bs, BdrvChild *ignore);
589 
590 /**
591  * bdrv_parent_drained_end:
592  *
593  * End a quiesced section of all users of @bs. This is part of
594  * bdrv_drained_end.
595  */
596 void bdrv_parent_drained_end(BlockDriverState *bs, BdrvChild *ignore);
597 
598 /**
599  * bdrv_drained_begin:
600  *
601  * Begin a quiesced section for exclusive access to the BDS, by disabling
602  * external request sources including NBD server and device model. Note that
603  * this doesn't block timers or coroutines from submitting more requests, which
604  * means block_job_pause is still necessary.
605  *
606  * This function can be recursive.
607  */
608 void bdrv_drained_begin(BlockDriverState *bs);
609 
610 /**
611  * Like bdrv_drained_begin, but recursively begins a quiesced section for
612  * exclusive access to all child nodes as well.
613  */
614 void bdrv_subtree_drained_begin(BlockDriverState *bs);
615 
616 /**
617  * bdrv_drained_end:
618  *
619  * End a quiescent section started by bdrv_drained_begin().
620  */
621 void bdrv_drained_end(BlockDriverState *bs);
622 
623 /**
624  * End a quiescent section started by bdrv_subtree_drained_begin().
625  */
626 void bdrv_subtree_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 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
633                                      uint32_t granularity, Error **errp);
634 /**
635  *
636  * bdrv_register_buf/bdrv_unregister_buf:
637  *
638  * Register/unregister a buffer for I/O. For example, VFIO drivers are
639  * interested to know the memory areas that would later be used for I/O, so
640  * that they can prepare IOMMU mapping etc., to get better performance.
641  */
642 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
643 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
644 #endif
645