xref: /openbmc/qemu/include/block/block.h (revision 1580b897)
1 #ifndef BLOCK_H
2 #define BLOCK_H
3 
4 #include "block/aio.h"
5 #include "block/aio-wait.h"
6 #include "qemu/iov.h"
7 #include "qemu/coroutine.h"
8 #include "block/accounting.h"
9 #include "block/dirty-bitmap.h"
10 #include "block/blockjob.h"
11 #include "qemu/hbitmap.h"
12 #include "qemu/transactions.h"
13 
14 /*
15  * generated_co_wrapper
16  *
17  * Function specifier, which does nothing but mark functions to be
18  * generated by scripts/block-coroutine-wrapper.py
19  *
20  * Read more in docs/devel/block-coroutine-wrapper.rst
21  */
22 #define generated_co_wrapper
23 
24 /* block.c */
25 typedef struct BlockDriver BlockDriver;
26 typedef struct BdrvChild BdrvChild;
27 typedef struct BdrvChildClass BdrvChildClass;
28 
29 typedef struct BlockDriverInfo {
30     /* in bytes, 0 if irrelevant */
31     int cluster_size;
32     /* offset at which the VM state can be saved (0 if not possible) */
33     int64_t vm_state_offset;
34     bool is_dirty;
35     /*
36      * True if this block driver only supports compressed writes
37      */
38     bool needs_compressed_writes;
39 } BlockDriverInfo;
40 
41 typedef struct BlockFragInfo {
42     uint64_t allocated_clusters;
43     uint64_t total_clusters;
44     uint64_t fragmented_clusters;
45     uint64_t compressed_clusters;
46 } BlockFragInfo;
47 
48 typedef enum {
49     BDRV_REQ_COPY_ON_READ       = 0x1,
50     BDRV_REQ_ZERO_WRITE         = 0x2,
51 
52     /*
53      * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
54      * that the block driver should unmap (discard) blocks if it is guaranteed
55      * that the result will read back as zeroes. The flag is only passed to the
56      * driver if the block device is opened with BDRV_O_UNMAP.
57      */
58     BDRV_REQ_MAY_UNMAP          = 0x4,
59 
60     BDRV_REQ_FUA                = 0x10,
61     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
62 
63     /* Signifies that this write request will not change the visible disk
64      * content. */
65     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
66 
67     /* Forces request serialisation. Use only with write requests. */
68     BDRV_REQ_SERIALISING        = 0x80,
69 
70     /* Execute the request only if the operation can be offloaded or otherwise
71      * be executed efficiently, but return an error instead of using a slow
72      * fallback. */
73     BDRV_REQ_NO_FALLBACK        = 0x100,
74 
75     /*
76      * BDRV_REQ_PREFETCH makes sense only in the context of copy-on-read
77      * (i.e., together with the BDRV_REQ_COPY_ON_READ flag or when a COR
78      * filter is involved), in which case it signals that the COR operation
79      * need not read the data into memory (qiov) but only ensure they are
80      * copied to the top layer (i.e., that COR operation is done).
81      */
82     BDRV_REQ_PREFETCH  = 0x200,
83 
84     /*
85      * If we need to wait for other requests, just fail immediately. Used
86      * only together with BDRV_REQ_SERIALISING.
87      */
88     BDRV_REQ_NO_WAIT = 0x400,
89 
90     /* Mask of valid flags */
91     BDRV_REQ_MASK               = 0x7ff,
92 } BdrvRequestFlags;
93 
94 typedef struct BlockSizes {
95     uint32_t phys;
96     uint32_t log;
97 } BlockSizes;
98 
99 typedef struct HDGeometry {
100     uint32_t heads;
101     uint32_t sectors;
102     uint32_t cylinders;
103 } HDGeometry;
104 
105 #define BDRV_O_NO_SHARE    0x0001 /* don't share permissions */
106 #define BDRV_O_RDWR        0x0002
107 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
108 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
109 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
110 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
111 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
112 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
113 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
114 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
115 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
116 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
117 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
118 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
119 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
120                                       select an appropriate protocol driver,
121                                       ignoring the format layer */
122 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
123 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
124 #define BDRV_O_IO_URING    0x40000 /* use io_uring instead of the thread pool */
125 
126 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
127 
128 
129 /* Option names of options parsed by the block layer */
130 
131 #define BDRV_OPT_CACHE_WB       "cache.writeback"
132 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
133 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
134 #define BDRV_OPT_READ_ONLY      "read-only"
135 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
136 #define BDRV_OPT_DISCARD        "discard"
137 #define BDRV_OPT_FORCE_SHARE    "force-share"
138 
139 
140 #define BDRV_SECTOR_BITS   9
141 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
142 
143 #define BDRV_REQUEST_MAX_SECTORS MIN_CONST(SIZE_MAX >> BDRV_SECTOR_BITS, \
144                                            INT_MAX >> BDRV_SECTOR_BITS)
145 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
146 
147 /*
148  * We want allow aligning requests and disk length up to any 32bit alignment
149  * and don't afraid of overflow.
150  * To achieve it, and in the same time use some pretty number as maximum disk
151  * size, let's define maximum "length" (a limit for any offset/bytes request and
152  * for disk size) to be the greatest power of 2 less than INT64_MAX.
153  */
154 #define BDRV_MAX_ALIGNMENT (1L << 30)
155 #define BDRV_MAX_LENGTH (QEMU_ALIGN_DOWN(INT64_MAX, BDRV_MAX_ALIGNMENT))
156 
157 /*
158  * Allocation status flags for bdrv_block_status() and friends.
159  *
160  * Public flags:
161  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
162  * BDRV_BLOCK_ZERO: offset reads as zero
163  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
164  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
165  *                       layer rather than any backing, set by block layer
166  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
167  *                 layer, set by block layer
168  *
169  * Internal flags:
170  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
171  *                 that the block layer recompute the answer from the returned
172  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
173  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
174  *                     zeroes in file child of current block node inside
175  *                     returned region. Only valid together with both
176  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
177  *                     appear with BDRV_BLOCK_ZERO.
178  *
179  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
180  * host offset within the returned BDS that is allocated for the
181  * corresponding raw guest data.  However, whether that offset
182  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
183  *
184  * DATA ZERO OFFSET_VALID
185  *  t    t        t       sectors read as zero, returned file is zero at offset
186  *  t    f        t       sectors read as valid from file at offset
187  *  f    t        t       sectors preallocated, read as zero, returned file not
188  *                        necessarily zero at offset
189  *  f    f        t       sectors preallocated but read from backing_hd,
190  *                        returned file contains garbage at offset
191  *  t    t        f       sectors preallocated, read as zero, unknown offset
192  *  t    f        f       sectors read from unknown file or offset
193  *  f    t        f       not allocated or unknown offset, read as zero
194  *  f    f        f       not allocated or unknown offset, read from backing_hd
195  */
196 #define BDRV_BLOCK_DATA         0x01
197 #define BDRV_BLOCK_ZERO         0x02
198 #define BDRV_BLOCK_OFFSET_VALID 0x04
199 #define BDRV_BLOCK_RAW          0x08
200 #define BDRV_BLOCK_ALLOCATED    0x10
201 #define BDRV_BLOCK_EOF          0x20
202 #define BDRV_BLOCK_RECURSE      0x40
203 
204 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
205 
206 typedef struct BDRVReopenState {
207     BlockDriverState *bs;
208     int flags;
209     BlockdevDetectZeroesOptions detect_zeroes;
210     bool backing_missing;
211     BlockDriverState *old_backing_bs; /* keep pointer for permissions update */
212     BlockDriverState *old_file_bs; /* keep pointer for permissions update */
213     QDict *options;
214     QDict *explicit_options;
215     void *opaque;
216 } BDRVReopenState;
217 
218 /*
219  * Block operation types
220  */
221 typedef enum BlockOpType {
222     BLOCK_OP_TYPE_BACKUP_SOURCE,
223     BLOCK_OP_TYPE_BACKUP_TARGET,
224     BLOCK_OP_TYPE_CHANGE,
225     BLOCK_OP_TYPE_COMMIT_SOURCE,
226     BLOCK_OP_TYPE_COMMIT_TARGET,
227     BLOCK_OP_TYPE_DATAPLANE,
228     BLOCK_OP_TYPE_DRIVE_DEL,
229     BLOCK_OP_TYPE_EJECT,
230     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
231     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
232     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
233     BLOCK_OP_TYPE_MIRROR_SOURCE,
234     BLOCK_OP_TYPE_MIRROR_TARGET,
235     BLOCK_OP_TYPE_RESIZE,
236     BLOCK_OP_TYPE_STREAM,
237     BLOCK_OP_TYPE_REPLACE,
238     BLOCK_OP_TYPE_MAX,
239 } BlockOpType;
240 
241 /* Block node permission constants */
242 enum {
243     /**
244      * A user that has the "permission" of consistent reads is guaranteed that
245      * their view of the contents of the block device is complete and
246      * self-consistent, representing the contents of a disk at a specific
247      * point.
248      *
249      * For most block devices (including their backing files) this is true, but
250      * the property cannot be maintained in a few situations like for
251      * intermediate nodes of a commit block job.
252      */
253     BLK_PERM_CONSISTENT_READ    = 0x01,
254 
255     /** This permission is required to change the visible disk contents. */
256     BLK_PERM_WRITE              = 0x02,
257 
258     /**
259      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
260      * required for writes to the block node when the caller promises that
261      * the visible disk content doesn't change.
262      *
263      * As the BLK_PERM_WRITE permission is strictly stronger, either is
264      * sufficient to perform an unchanging write.
265      */
266     BLK_PERM_WRITE_UNCHANGED    = 0x04,
267 
268     /** This permission is required to change the size of a block node. */
269     BLK_PERM_RESIZE             = 0x08,
270 
271     /**
272      * This permission is required to change the node that this BdrvChild
273      * points to.
274      */
275     BLK_PERM_GRAPH_MOD          = 0x10,
276 
277     BLK_PERM_ALL                = 0x1f,
278 
279     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
280                                  | BLK_PERM_WRITE
281                                  | BLK_PERM_WRITE_UNCHANGED
282                                  | BLK_PERM_RESIZE,
283 
284     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
285 };
286 
287 /*
288  * Flags that parent nodes assign to child nodes to specify what kind of
289  * role(s) they take.
290  *
291  * At least one of DATA, METADATA, FILTERED, or COW must be set for
292  * every child.
293  */
294 enum BdrvChildRoleBits {
295     /*
296      * This child stores data.
297      * Any node may have an arbitrary number of such children.
298      */
299     BDRV_CHILD_DATA         = (1 << 0),
300 
301     /*
302      * This child stores metadata.
303      * Any node may have an arbitrary number of metadata-storing
304      * children.
305      */
306     BDRV_CHILD_METADATA     = (1 << 1),
307 
308     /*
309      * A child that always presents exactly the same visible data as
310      * the parent, e.g. by virtue of the parent forwarding all reads
311      * and writes.
312      * This flag is mutually exclusive with DATA, METADATA, and COW.
313      * Any node may have at most one filtered child at a time.
314      */
315     BDRV_CHILD_FILTERED     = (1 << 2),
316 
317     /*
318      * Child from which to read all data that isn't allocated in the
319      * parent (i.e., the backing child); such data is copied to the
320      * parent through COW (and optionally COR).
321      * This field is mutually exclusive with DATA, METADATA, and
322      * FILTERED.
323      * Any node may have at most one such backing child at a time.
324      */
325     BDRV_CHILD_COW          = (1 << 3),
326 
327     /*
328      * The primary child.  For most drivers, this is the child whose
329      * filename applies best to the parent node.
330      * Any node may have at most one primary child at a time.
331      */
332     BDRV_CHILD_PRIMARY      = (1 << 4),
333 
334     /* Useful combination of flags */
335     BDRV_CHILD_IMAGE        = BDRV_CHILD_DATA
336                               | BDRV_CHILD_METADATA
337                               | BDRV_CHILD_PRIMARY,
338 };
339 
340 /* Mask of BdrvChildRoleBits values */
341 typedef unsigned int BdrvChildRole;
342 
343 char *bdrv_perm_names(uint64_t perm);
344 uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm);
345 
346 /* disk I/O throttling */
347 void bdrv_init(void);
348 void bdrv_init_with_whitelist(void);
349 bool bdrv_uses_whitelist(void);
350 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
351 BlockDriver *bdrv_find_protocol(const char *filename,
352                                 bool allow_protocol_prefix,
353                                 Error **errp);
354 BlockDriver *bdrv_find_format(const char *format_name);
355 int bdrv_create(BlockDriver *drv, const char* filename,
356                 QemuOpts *opts, Error **errp);
357 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
358 
359 BlockDriverState *bdrv_new(void);
360 int bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
361                 Error **errp);
362 int bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
363                       Error **errp);
364 BlockDriverState *bdrv_insert_node(BlockDriverState *bs, QDict *node_options,
365                                    int flags, Error **errp);
366 int bdrv_drop_filter(BlockDriverState *bs, Error **errp);
367 
368 int bdrv_parse_aio(const char *mode, int *flags);
369 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
370 int bdrv_parse_discard_flags(const char *mode, int *flags);
371 BdrvChild *bdrv_open_child(const char *filename,
372                            QDict *options, const char *bdref_key,
373                            BlockDriverState* parent,
374                            const BdrvChildClass *child_class,
375                            BdrvChildRole child_role,
376                            bool allow_none, Error **errp);
377 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
378 int bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
379                         Error **errp);
380 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
381                            const char *bdref_key, Error **errp);
382 BlockDriverState *bdrv_open(const char *filename, const char *reference,
383                             QDict *options, int flags, Error **errp);
384 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
385                                        int flags, Error **errp);
386 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
387                                     BlockDriverState *bs, QDict *options,
388                                     bool keep_old_opts);
389 void bdrv_reopen_queue_free(BlockReopenQueue *bs_queue);
390 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
391 int bdrv_reopen(BlockDriverState *bs, QDict *opts, bool keep_old_opts,
392                 Error **errp);
393 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
394                               Error **errp);
395 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
396                        int64_t bytes, BdrvRequestFlags flags);
397 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
398 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int64_t bytes);
399 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf,
400                 int64_t bytes);
401 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
402                      const void *buf, int64_t bytes);
403 /*
404  * Efficiently zero a region of the disk image.  Note that this is a regular
405  * I/O request like read or write and should have a reasonable size.  This
406  * function is not suitable for zeroing the entire image in a single request
407  * because it may allocate memory for the entire region.
408  */
409 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
410                                        int64_t bytes, BdrvRequestFlags flags);
411 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
412     const char *backing_file);
413 void bdrv_refresh_filename(BlockDriverState *bs);
414 
415 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
416                                   PreallocMode prealloc, BdrvRequestFlags flags,
417                                   Error **errp);
418 int generated_co_wrapper
419 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
420               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
421 
422 int64_t bdrv_nb_sectors(BlockDriverState *bs);
423 int64_t bdrv_getlength(BlockDriverState *bs);
424 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
425 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
426                                BlockDriverState *in_bs, Error **errp);
427 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
428 void bdrv_refresh_limits(BlockDriverState *bs, Transaction *tran, Error **errp);
429 int bdrv_commit(BlockDriverState *bs);
430 int bdrv_make_empty(BdrvChild *c, Error **errp);
431 int bdrv_change_backing_file(BlockDriverState *bs, const char *backing_file,
432                              const char *backing_fmt, bool warn);
433 void bdrv_register(BlockDriver *bdrv);
434 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
435                            const char *backing_file_str);
436 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
437                                     BlockDriverState *bs);
438 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
439 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
440                                   Error **errp);
441 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
442                               Error **errp);
443 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
444 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
445 void coroutine_fn bdrv_co_delete_file_noerr(BlockDriverState *bs);
446 
447 
448 typedef struct BdrvCheckResult {
449     int corruptions;
450     int leaks;
451     int check_errors;
452     int corruptions_fixed;
453     int leaks_fixed;
454     int64_t image_end_offset;
455     BlockFragInfo bfi;
456 } BdrvCheckResult;
457 
458 typedef enum {
459     BDRV_FIX_LEAKS    = 1,
460     BDRV_FIX_ERRORS   = 2,
461 } BdrvCheckMode;
462 
463 int generated_co_wrapper bdrv_check(BlockDriverState *bs, BdrvCheckResult *res,
464                                     BdrvCheckMode fix);
465 
466 /* The units of offset and total_work_size may be chosen arbitrarily by the
467  * block driver; total_work_size may change during the course of the amendment
468  * operation */
469 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
470                                       int64_t total_work_size, void *opaque);
471 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
472                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
473                        bool force,
474                        Error **errp);
475 
476 /* check if a named node can be replaced when doing drive-mirror */
477 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
478                                         const char *node_name, Error **errp);
479 
480 /* async block I/O */
481 void bdrv_aio_cancel(BlockAIOCB *acb);
482 void bdrv_aio_cancel_async(BlockAIOCB *acb);
483 
484 /* sg packet commands */
485 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
486 
487 /* Invalidate any cached metadata used by image formats */
488 int generated_co_wrapper bdrv_invalidate_cache(BlockDriverState *bs,
489                                                Error **errp);
490 void bdrv_invalidate_cache_all(Error **errp);
491 int bdrv_inactivate_all(void);
492 
493 /* Ensure contents are flushed to disk.  */
494 int generated_co_wrapper bdrv_flush(BlockDriverState *bs);
495 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
496 int bdrv_flush_all(void);
497 void bdrv_close_all(void);
498 void bdrv_drain(BlockDriverState *bs);
499 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
500 void bdrv_drain_all_begin(void);
501 void bdrv_drain_all_end(void);
502 void bdrv_drain_all(void);
503 
504 #define BDRV_POLL_WHILE(bs, cond) ({                       \
505     BlockDriverState *bs_ = (bs);                          \
506     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
507                    cond); })
508 
509 int generated_co_wrapper bdrv_pdiscard(BdrvChild *child, int64_t offset,
510                                        int64_t bytes);
511 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
512 int bdrv_has_zero_init_1(BlockDriverState *bs);
513 int bdrv_has_zero_init(BlockDriverState *bs);
514 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
515 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
516                       int64_t bytes, int64_t *pnum, int64_t *map,
517                       BlockDriverState **file);
518 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
519                             int64_t offset, int64_t bytes, int64_t *pnum,
520                             int64_t *map, BlockDriverState **file);
521 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
522                       int64_t *pnum);
523 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
524                             bool include_base, int64_t offset, int64_t bytes,
525                             int64_t *pnum);
526 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset,
527                                       int64_t bytes);
528 
529 bool bdrv_is_read_only(BlockDriverState *bs);
530 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
531                            bool ignore_allow_rdw, Error **errp);
532 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
533                               Error **errp);
534 bool bdrv_is_writable(BlockDriverState *bs);
535 bool bdrv_is_sg(BlockDriverState *bs);
536 bool bdrv_is_inserted(BlockDriverState *bs);
537 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
538 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
539 const char *bdrv_get_format_name(BlockDriverState *bs);
540 BlockDriverState *bdrv_find_node(const char *node_name);
541 BlockDeviceInfoList *bdrv_named_nodes_list(bool flat, Error **errp);
542 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
543 BlockDriverState *bdrv_lookup_bs(const char *device,
544                                  const char *node_name,
545                                  Error **errp);
546 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
547 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
548 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
549 
550 typedef struct BdrvNextIterator {
551     enum {
552         BDRV_NEXT_BACKEND_ROOTS,
553         BDRV_NEXT_MONITOR_OWNED,
554     } phase;
555     BlockBackend *blk;
556     BlockDriverState *bs;
557 } BdrvNextIterator;
558 
559 BlockDriverState *bdrv_first(BdrvNextIterator *it);
560 BlockDriverState *bdrv_next(BdrvNextIterator *it);
561 void bdrv_next_cleanup(BdrvNextIterator *it);
562 
563 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
564 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
565 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
566                          void *opaque, bool read_only);
567 const char *bdrv_get_node_name(const BlockDriverState *bs);
568 const char *bdrv_get_device_name(const BlockDriverState *bs);
569 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
570 int bdrv_get_flags(BlockDriverState *bs);
571 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
572 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
573                                           Error **errp);
574 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
575 void bdrv_round_to_clusters(BlockDriverState *bs,
576                             int64_t offset, int64_t bytes,
577                             int64_t *cluster_offset,
578                             int64_t *cluster_bytes);
579 
580 void bdrv_get_backing_filename(BlockDriverState *bs,
581                                char *filename, int filename_size);
582 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
583 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
584                                                    const char *backing,
585                                                    Error **errp);
586 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
587 
588 int path_has_protocol(const char *path);
589 int path_is_absolute(const char *path);
590 char *path_combine(const char *base_path, const char *filename);
591 
592 int generated_co_wrapper
593 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
594 int generated_co_wrapper
595 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
596 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
597                       int64_t pos, int size);
598 
599 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
600                       int64_t pos, int size);
601 
602 void bdrv_img_create(const char *filename, const char *fmt,
603                      const char *base_filename, const char *base_fmt,
604                      char *options, uint64_t img_size, int flags,
605                      bool quiet, Error **errp);
606 
607 /* Returns the alignment in bytes that is required so that no bounce buffer
608  * is required throughout the stack */
609 size_t bdrv_min_mem_align(BlockDriverState *bs);
610 /* Returns optimal alignment in bytes for bounce buffer */
611 size_t bdrv_opt_mem_align(BlockDriverState *bs);
612 void *qemu_blockalign(BlockDriverState *bs, size_t size);
613 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
614 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
615 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
616 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
617 
618 void bdrv_enable_copy_on_read(BlockDriverState *bs);
619 void bdrv_disable_copy_on_read(BlockDriverState *bs);
620 
621 void bdrv_ref(BlockDriverState *bs);
622 void bdrv_unref(BlockDriverState *bs);
623 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
624 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
625                              BlockDriverState *child_bs,
626                              const char *child_name,
627                              const BdrvChildClass *child_class,
628                              BdrvChildRole child_role,
629                              Error **errp);
630 
631 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
632 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
633 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
634 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
635 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
636 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
637 
638 #define BLKDBG_EVENT(child, evt) \
639     do { \
640         if (child) { \
641             bdrv_debug_event(child->bs, evt); \
642         } \
643     } while (0)
644 
645 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
646 
647 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
648                            const char *tag);
649 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
650 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
651 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
652 
653 /**
654  * bdrv_get_aio_context:
655  *
656  * Returns: the currently bound #AioContext
657  */
658 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
659 
660 /**
661  * Move the current coroutine to the AioContext of @bs and return the old
662  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
663  * will wait for the operation to proceed until the corresponding
664  * bdrv_co_leave().
665  *
666  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
667  * this will deadlock.
668  */
669 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
670 
671 /**
672  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
673  * to old_ctx and decrease bs->in_flight again.
674  */
675 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
676 
677 /**
678  * Locks the AioContext of @bs if it's not the current AioContext. This avoids
679  * double locking which could lead to deadlocks: This is a coroutine_fn, so we
680  * know we already own the lock of the current AioContext.
681  *
682  * May only be called in the main thread.
683  */
684 void coroutine_fn bdrv_co_lock(BlockDriverState *bs);
685 
686 /**
687  * Unlocks the AioContext of @bs if it's not the current AioContext.
688  */
689 void coroutine_fn bdrv_co_unlock(BlockDriverState *bs);
690 
691 /**
692  * Transfer control to @co in the aio context of @bs
693  */
694 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
695 
696 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
697                                  AioContext *new_context, GSList **ignore);
698 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
699                              Error **errp);
700 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
701                                    BdrvChild *ignore_child, Error **errp);
702 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
703                                     GSList **ignore, Error **errp);
704 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
705                               GSList **ignore, Error **errp);
706 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c);
707 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c);
708 
709 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
710 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
711 
712 void bdrv_io_plug(BlockDriverState *bs);
713 void bdrv_io_unplug(BlockDriverState *bs);
714 
715 /**
716  * bdrv_parent_drained_begin_single:
717  *
718  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
719  * any pending activity to cease.
720  */
721 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
722 
723 /**
724  * bdrv_parent_drained_end_single:
725  *
726  * End a quiesced section for the parent of @c.
727  *
728  * This polls @bs's AioContext until all scheduled sub-drained_ends
729  * have settled, which may result in graph changes.
730  */
731 void bdrv_parent_drained_end_single(BdrvChild *c);
732 
733 /**
734  * bdrv_drain_poll:
735  *
736  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
737  * and if @recursive is true its children as well (used for subtree drain).
738  *
739  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
740  * ignore the drain request because they will be drained separately (used for
741  * drain_all).
742  *
743  * This is part of bdrv_drained_begin.
744  */
745 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
746                      BdrvChild *ignore_parent, bool ignore_bds_parents);
747 
748 /**
749  * bdrv_drained_begin:
750  *
751  * Begin a quiesced section for exclusive access to the BDS, by disabling
752  * external request sources including NBD server and device model. Note that
753  * this doesn't block timers or coroutines from submitting more requests, which
754  * means block_job_pause is still necessary.
755  *
756  * This function can be recursive.
757  */
758 void bdrv_drained_begin(BlockDriverState *bs);
759 
760 /**
761  * bdrv_do_drained_begin_quiesce:
762  *
763  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
764  * running requests to complete.
765  */
766 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
767                                    BdrvChild *parent, bool ignore_bds_parents);
768 
769 /**
770  * Like bdrv_drained_begin, but recursively begins a quiesced section for
771  * exclusive access to all child nodes as well.
772  */
773 void bdrv_subtree_drained_begin(BlockDriverState *bs);
774 
775 /**
776  * bdrv_drained_end:
777  *
778  * End a quiescent section started by bdrv_drained_begin().
779  *
780  * This polls @bs's AioContext until all scheduled sub-drained_ends
781  * have settled.  On one hand, that may result in graph changes.  On
782  * the other, this requires that the caller either runs in the main
783  * loop; or that all involved nodes (@bs and all of its parents) are
784  * in the caller's AioContext.
785  */
786 void bdrv_drained_end(BlockDriverState *bs);
787 
788 /**
789  * bdrv_drained_end_no_poll:
790  *
791  * Same as bdrv_drained_end(), but do not poll for the subgraph to
792  * actually become unquiesced.  Therefore, no graph changes will occur
793  * with this function.
794  *
795  * *drained_end_counter is incremented for every background operation
796  * that is scheduled, and will be decremented for every operation once
797  * it settles.  The caller must poll until it reaches 0.  The counter
798  * should be accessed using atomic operations only.
799  */
800 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
801 
802 /**
803  * End a quiescent section started by bdrv_subtree_drained_begin().
804  */
805 void bdrv_subtree_drained_end(BlockDriverState *bs);
806 
807 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
808                     Error **errp);
809 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
810 
811 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
812                                      uint32_t granularity, Error **errp);
813 /**
814  *
815  * bdrv_register_buf/bdrv_unregister_buf:
816  *
817  * Register/unregister a buffer for I/O. For example, VFIO drivers are
818  * interested to know the memory areas that would later be used for I/O, so
819  * that they can prepare IOMMU mapping etc., to get better performance.
820  */
821 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
822 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
823 
824 /**
825  *
826  * bdrv_co_copy_range:
827  *
828  * Do offloaded copy between two children. If the operation is not implemented
829  * by the driver, or if the backend storage doesn't support it, a negative
830  * error code will be returned.
831  *
832  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
833  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
834  * calling copy_file_range(2)) after the first error, thus it should fall back
835  * to a read+write path in the caller level.
836  *
837  * @src: Source child to copy data from
838  * @src_offset: offset in @src image to read data
839  * @dst: Destination child to copy data to
840  * @dst_offset: offset in @dst image to write data
841  * @bytes: number of bytes to copy
842  * @flags: request flags. Supported flags:
843  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
844  *                               write on @dst as if bdrv_co_pwrite_zeroes is
845  *                               called. Used to simplify caller code, or
846  *                               during BlockDriver.bdrv_co_copy_range_from()
847  *                               recursion.
848  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
849  *                                   requests currently in flight.
850  *
851  * Returns: 0 if succeeded; negative error code if failed.
852  **/
853 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, int64_t src_offset,
854                                     BdrvChild *dst, int64_t dst_offset,
855                                     int64_t bytes, BdrvRequestFlags read_flags,
856                                     BdrvRequestFlags write_flags);
857 
858 void bdrv_cancel_in_flight(BlockDriverState *bs);
859 
860 #endif
861