]> git.proxmox.com Git - mirror_qemu.git/blame - include/block/block_int-common.h
block/block-common: add zoned device structs
[mirror_qemu.git] / include / block / block_int-common.h
CommitLineData
ebc2752b
EGE
1/*
2 * QEMU System Emulator block driver
3 *
4 * Copyright (c) 2003 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24#ifndef BLOCK_INT_COMMON_H
25#define BLOCK_INT_COMMON_H
26
e2c1c34f
MA
27#include "block/aio.h"
28#include "block/block-common.h"
29#include "block/block-global-state.h"
ebc2752b 30#include "block/snapshot.h"
e2c1c34f 31#include "qemu/iov.h"
ebc2752b 32#include "qemu/rcu.h"
e2c1c34f 33#include "qemu/stats64.h"
ebc2752b
EGE
34
35#define BLOCK_FLAG_LAZY_REFCOUNTS 8
36
37#define BLOCK_OPT_SIZE "size"
38#define BLOCK_OPT_ENCRYPT "encryption"
39#define BLOCK_OPT_ENCRYPT_FORMAT "encrypt.format"
40#define BLOCK_OPT_COMPAT6 "compat6"
41#define BLOCK_OPT_HWVERSION "hwversion"
42#define BLOCK_OPT_BACKING_FILE "backing_file"
43#define BLOCK_OPT_BACKING_FMT "backing_fmt"
44#define BLOCK_OPT_CLUSTER_SIZE "cluster_size"
45#define BLOCK_OPT_TABLE_SIZE "table_size"
46#define BLOCK_OPT_PREALLOC "preallocation"
47#define BLOCK_OPT_SUBFMT "subformat"
48#define BLOCK_OPT_COMPAT_LEVEL "compat"
49#define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts"
50#define BLOCK_OPT_ADAPTER_TYPE "adapter_type"
51#define BLOCK_OPT_REDUNDANCY "redundancy"
52#define BLOCK_OPT_NOCOW "nocow"
53#define BLOCK_OPT_EXTENT_SIZE_HINT "extent_size_hint"
54#define BLOCK_OPT_OBJECT_SIZE "object_size"
55#define BLOCK_OPT_REFCOUNT_BITS "refcount_bits"
56#define BLOCK_OPT_DATA_FILE "data_file"
57#define BLOCK_OPT_DATA_FILE_RAW "data_file_raw"
58#define BLOCK_OPT_COMPRESSION_TYPE "compression_type"
59#define BLOCK_OPT_EXTL2 "extended_l2"
60
61#define BLOCK_PROBE_BUF_SIZE 512
62
63enum BdrvTrackedRequestType {
64 BDRV_TRACKED_READ,
65 BDRV_TRACKED_WRITE,
66 BDRV_TRACKED_DISCARD,
67 BDRV_TRACKED_TRUNCATE,
68};
69
70/*
71 * That is not quite good that BdrvTrackedRequest structure is public,
72 * as block/io.c is very careful about incoming offset/bytes being
73 * correct. Be sure to assert bdrv_check_request() succeeded after any
74 * modification of BdrvTrackedRequest object out of block/io.c
75 */
76typedef struct BdrvTrackedRequest {
77 BlockDriverState *bs;
78 int64_t offset;
79 int64_t bytes;
80 enum BdrvTrackedRequestType type;
81
82 bool serialising;
83 int64_t overlap_offset;
84 int64_t overlap_bytes;
85
86 QLIST_ENTRY(BdrvTrackedRequest) list;
87 Coroutine *co; /* owner, used for deadlock detection */
88 CoQueue wait_queue; /* coroutines blocked on this request */
89
90 struct BdrvTrackedRequest *waiting_for;
91} BdrvTrackedRequest;
92
93
94struct BlockDriver {
69c0bf11
EGE
95 /*
96 * These fields are initialized when this object is created,
97 * and are never changed afterwards.
98 */
99
ebc2752b
EGE
100 const char *format_name;
101 int instance_size;
102
103 /*
104 * Set to true if the BlockDriver is a block filter. Block filters pass
105 * certain callbacks that refer to data (see block.c) to their bs->file
106 * or bs->backing (whichever one exists) if the driver doesn't implement
107 * them. Drivers that do not wish to forward must implement them and return
108 * -ENOTSUP.
109 * Note that filters are not allowed to modify data.
110 *
111 * Filters generally cannot have more than a single filtered child,
112 * because the data they present must at all times be the same as
113 * that on their filtered child. That would be impossible to
114 * achieve for multiple filtered children.
115 * (And this filtered child must then be bs->file or bs->backing.)
116 */
117 bool is_filter;
046fd84f
VSO
118 /*
119 * Only make sense for filter drivers, for others must be false.
120 * If true, filtered child is bs->backing. Otherwise it's bs->file.
1921b4f7
VSO
121 * Two internal filters use bs->backing as filtered child and has this
122 * field set to true: mirror_top and commit_top. There also two such test
123 * filters in tests/unit/test-bdrv-graph-mod.c.
046fd84f
VSO
124 *
125 * Never create any more such filters!
126 *
127 * TODO: imagine how to deprecate this behavior and make all filters work
128 * similarly using bs->file as filtered child.
129 */
130 bool filtered_child_is_backing;
131
ebc2752b
EGE
132 /*
133 * Set to true if the BlockDriver is a format driver. Format nodes
134 * generally do not expect their children to be other format nodes
135 * (except for backing files), and so format probing is disabled
136 * on those children.
137 */
138 bool is_format;
139
69c0bf11
EGE
140 /*
141 * Drivers not implementing bdrv_parse_filename nor bdrv_open should have
142 * this field set to true, except ones that are defined only by their
143 * child's bs.
144 * An example of the last type will be the quorum block driver.
145 */
146 bool bdrv_needs_filename;
147
148 /*
149 * Set if a driver can support backing files. This also implies the
150 * following semantics:
151 *
152 * - Return status 0 of .bdrv_co_block_status means that corresponding
153 * blocks are not allocated in this layer of backing-chain
154 * - For such (unallocated) blocks, read will:
155 * - fill buffer with zeros if there is no backing file
156 * - read from the backing file otherwise, where the block layer
157 * takes care of reading zeros beyond EOF if backing file is short
158 */
159 bool supports_backing;
160
69c0bf11
EGE
161 /*
162 * Drivers setting this field must be able to work with just a plain
163 * filename with '<protocol_name>:' as a prefix, and no other options.
164 * Options may be extracted from the filename by implementing
165 * bdrv_parse_filename.
166 */
167 const char *protocol_name;
168
169 /* List of options for creating images, terminated by name == NULL */
170 QemuOptsList *create_opts;
171
172 /* List of options for image amend */
173 QemuOptsList *amend_opts;
174
175 /*
176 * If this driver supports reopening images this contains a
177 * NULL-terminated list of the runtime options that can be
178 * modified. If an option in this list is unspecified during
179 * reopen then it _must_ be reset to its default value or return
180 * an error.
181 */
182 const char *const *mutable_opts;
183
184 /*
185 * Pointer to a NULL-terminated array of names of strong options
186 * that can be specified for bdrv_open(). A strong option is one
187 * that changes the data of a BDS.
188 * If this pointer is NULL, the array is considered empty.
189 * "filename" and "driver" are always considered strong.
190 */
191 const char *const *strong_runtime_opts;
192
193
194 /*
195 * Global state (GS) API. These functions run under the BQL.
196 *
197 * See include/block/block-global-state.h for more information about
198 * the GS API.
199 */
200
ebc2752b
EGE
201 /*
202 * This function is invoked under BQL before .bdrv_co_amend()
203 * (which in contrast does not necessarily run under the BQL)
204 * to allow driver-specific initialization code that requires
205 * the BQL, like setting up specific permission flags.
206 */
840428a2
EGE
207 int GRAPH_RDLOCK_PTR (*bdrv_amend_pre_run)(
208 BlockDriverState *bs, Error **errp);
ebc2752b
EGE
209 /*
210 * This function is invoked under BQL after .bdrv_co_amend()
211 * to allow cleaning up what was done in .bdrv_amend_pre_run().
212 */
840428a2 213 void GRAPH_RDLOCK_PTR (*bdrv_amend_clean)(BlockDriverState *bs);
ebc2752b
EGE
214
215 /*
216 * Return true if @to_replace can be replaced by a BDS with the
217 * same data as @bs without it affecting @bs's behavior (that is,
218 * without it being visible to @bs's parents).
219 */
533c6e4e
KW
220 bool GRAPH_RDLOCK_PTR (*bdrv_recurse_can_replace)(
221 BlockDriverState *bs, BlockDriverState *to_replace);
ebc2752b 222
ebc2752b
EGE
223 int (*bdrv_probe_device)(const char *filename);
224
225 /*
226 * Any driver implementing this callback is expected to be able to handle
227 * NULL file names in its .bdrv_open() implementation.
228 */
229 void (*bdrv_parse_filename)(const char *filename, QDict *options,
230 Error **errp);
ebc2752b 231
69c0bf11 232 /* For handling image reopen for split or non-split files. */
ebc2752b
EGE
233 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
234 BlockReopenQueue *queue, Error **errp);
235 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
236 void (*bdrv_reopen_commit_post)(BDRVReopenState *reopen_state);
237 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
238 void (*bdrv_join_options)(QDict *options, QDict *old_options);
239
1a30b0f5
KW
240 int GRAPH_UNLOCKED_PTR (*bdrv_open)(
241 BlockDriverState *bs, QDict *options, int flags, Error **errp);
ebc2752b
EGE
242
243 /* Protocol drivers should implement this instead of bdrv_open */
1a30b0f5
KW
244 int GRAPH_UNLOCKED_PTR (*bdrv_file_open)(
245 BlockDriverState *bs, QDict *options, int flags, Error **errp);
ebc2752b
EGE
246 void (*bdrv_close)(BlockDriverState *bs);
247
4ec8df01
KW
248 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_create)(
249 BlockdevCreateOptions *opts, Error **errp);
250
251 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_create_opts)(
252 BlockDriver *drv, const char *filename, QemuOpts *opts, Error **errp);
ebc2752b 253
ebc2752b
EGE
254 int (*bdrv_amend_options)(BlockDriverState *bs,
255 QemuOpts *opts,
256 BlockDriverAmendStatusCB *status_cb,
257 void *cb_opaque,
258 bool force,
259 Error **errp);
260
261 int (*bdrv_make_empty)(BlockDriverState *bs);
262
263 /*
264 * Refreshes the bs->exact_filename field. If that is impossible,
265 * bs->exact_filename has to be left empty.
266 */
267 void (*bdrv_refresh_filename)(BlockDriverState *bs);
268
269 /*
270 * Gathers the open options for all children into @target.
271 * A simple format driver (without backing file support) might
272 * implement this function like this:
273 *
274 * QINCREF(bs->file->bs->full_open_options);
275 * qdict_put(target, "file", bs->file->bs->full_open_options);
276 *
277 * If not specified, the generic implementation will simply put
278 * all children's options under their respective name.
279 *
280 * @backing_overridden is true when bs->backing seems not to be
281 * the child that would result from opening bs->backing_file.
282 * Therefore, if it is true, the backing child's options should be
283 * gathered; otherwise, there is no need since the backing child
284 * is the one implied by the image header.
285 *
286 * Note that ideally this function would not be needed. Every
287 * block driver which implements it is probably doing something
288 * shady regarding its runtime option structure.
289 */
290 void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target,
291 bool backing_overridden);
292
293 /*
294 * Returns an allocated string which is the directory name of this BDS: It
295 * will be used to make relative filenames absolute by prepending this
296 * function's return value to them.
297 */
298 char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp);
299
69c0bf11
EGE
300 /*
301 * This informs the driver that we are no longer interested in the result
302 * of in-flight requests, so don't waste the time if possible.
303 *
304 * One example usage is to avoid waiting for an nbd target node reconnect
305 * timeout during job-cancel with force=true.
306 */
307 void (*bdrv_cancel_in_flight)(BlockDriverState *bs);
308
309 int (*bdrv_inactivate)(BlockDriverState *bs);
310
311 int (*bdrv_snapshot_create)(BlockDriverState *bs,
312 QEMUSnapshotInfo *sn_info);
313 int (*bdrv_snapshot_goto)(BlockDriverState *bs,
314 const char *snapshot_id);
315 int (*bdrv_snapshot_delete)(BlockDriverState *bs,
316 const char *snapshot_id,
317 const char *name,
318 Error **errp);
319 int (*bdrv_snapshot_list)(BlockDriverState *bs,
320 QEMUSnapshotInfo **psn_info);
321 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
322 const char *snapshot_id,
323 const char *name,
324 Error **errp);
325
326 int (*bdrv_change_backing_file)(BlockDriverState *bs,
327 const char *backing_file, const char *backing_fmt);
328
329 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
330 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
331 const char *tag);
332 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
333 const char *tag);
334 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
335 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
336
e19b157f
KW
337 void GRAPH_RDLOCK_PTR (*bdrv_refresh_limits)(
338 BlockDriverState *bs, Error **errp);
69c0bf11
EGE
339
340 /*
341 * Returns 1 if newly created images are guaranteed to contain only
342 * zeros, 0 otherwise.
343 */
344 int (*bdrv_has_zero_init)(BlockDriverState *bs);
345
346 /*
347 * Remove fd handlers, timers, and other event loop callbacks so the event
348 * loop is no longer in use. Called with no in-flight requests and in
349 * depth-first traversal order with parents before child nodes.
350 */
351 void (*bdrv_detach_aio_context)(BlockDriverState *bs);
352
353 /*
354 * Add fd handlers, timers, and other event loop callbacks so I/O requests
355 * can be processed again. Called with no in-flight requests and in
356 * depth-first traversal order with child nodes before parent nodes.
357 */
358 void (*bdrv_attach_aio_context)(BlockDriverState *bs,
359 AioContext *new_context);
360
361 /**
362 * Try to get @bs's logical and physical block size.
363 * On success, store them in @bsz and return zero.
364 * On failure, return negative errno.
365 */
366 int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
367 /**
368 * Try to get @bs's geometry (cyls, heads, sectors)
369 * On success, store them in @geo and return 0.
370 * On failure return -errno.
371 * Only drivers that want to override guest geometry implement this
372 * callback; see hd_geometry_guess().
373 */
374 int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
375
376 void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
377 Error **errp);
378 void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
379 Error **errp);
380
381 /**
382 * Informs the block driver that a permission change is intended. The
383 * driver checks whether the change is permissible and may take other
384 * preparations for the change (e.g. get file system locks). This operation
385 * is always followed either by a call to either .bdrv_set_perm or
386 * .bdrv_abort_perm_update.
387 *
388 * Checks whether the requested set of cumulative permissions in @perm
389 * can be granted for accessing @bs and whether no other users are using
390 * permissions other than those given in @shared (both arguments take
391 * BLK_PERM_* bitmasks).
392 *
393 * If both conditions are met, 0 is returned. Otherwise, -errno is returned
394 * and errp is set to an error describing the conflict.
395 */
396 int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
397 uint64_t shared, Error **errp);
398
399 /**
400 * Called to inform the driver that the set of cumulative set of used
401 * permissions for @bs has changed to @perm, and the set of sharable
402 * permission to @shared. The driver can use this to propagate changes to
403 * its children (i.e. request permissions only if a parent actually needs
404 * them).
405 *
406 * This function is only invoked after bdrv_check_perm(), so block drivers
407 * may rely on preparations made in their .bdrv_check_perm implementation.
408 */
409 void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
410
411 /*
412 * Called to inform the driver that after a previous bdrv_check_perm()
413 * call, the permission update is not performed and any preparations made
414 * for it (e.g. taken file locks) need to be undone.
415 *
416 * This function can be called even for nodes that never saw a
417 * bdrv_check_perm() call. It is a no-op then.
418 */
419 void (*bdrv_abort_perm_update)(BlockDriverState *bs);
420
421 /**
422 * Returns in @nperm and @nshared the permissions that the driver for @bs
423 * needs on its child @c, based on the cumulative permissions requested by
424 * the parents in @parent_perm and @parent_shared.
425 *
426 * If @c is NULL, return the permissions for attaching a new child for the
427 * given @child_class and @role.
428 *
429 * If @reopen_queue is non-NULL, don't return the currently needed
430 * permissions, but those that will be needed after applying the
431 * @reopen_queue.
432 */
433 void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
434 BdrvChildRole role,
435 BlockReopenQueue *reopen_queue,
436 uint64_t parent_perm, uint64_t parent_shared,
437 uint64_t *nperm, uint64_t *nshared);
438
439 /**
440 * Register/unregister a buffer for I/O. For example, when the driver is
441 * interested to know the memory areas that will later be used in iovs, so
442 * that it can do IOMMU mapping with VFIO etc., in order to get better
443 * performance. In the case of VFIO drivers, this callback is used to do
444 * DMA mapping for hot buffers.
f4ec04ba
SH
445 *
446 * Returns: true on success, false on failure
69c0bf11 447 */
d9249c25
KW
448 bool GRAPH_RDLOCK_PTR (*bdrv_register_buf)(
449 BlockDriverState *bs, void *host, size_t size, Error **errp);
450 void GRAPH_RDLOCK_PTR (*bdrv_unregister_buf)(
451 BlockDriverState *bs, void *host, size_t size);
69c0bf11
EGE
452
453 /*
454 * This field is modified only under the BQL, and is part of
455 * the global state.
456 */
457 QLIST_ENTRY(BlockDriver) list;
458
459 /*
460 * I/O API functions. These functions are thread-safe.
461 *
462 * See include/block/block-io.h for more information about
463 * the I/O API.
464 */
465
466 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
467
840428a2
EGE
468 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_amend)(
469 BlockDriverState *bs, BlockdevAmendOptions *opts, bool force,
470 Error **errp);
69c0bf11 471
ebc2752b 472 /* aio */
7b1fb72e 473 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_preadv)(BlockDriverState *bs,
ebc2752b
EGE
474 int64_t offset, int64_t bytes, QEMUIOVector *qiov,
475 BdrvRequestFlags flags, BlockCompletionFunc *cb, void *opaque);
7b1fb72e
KW
476
477 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_pwritev)(BlockDriverState *bs,
ebc2752b
EGE
478 int64_t offset, int64_t bytes, QEMUIOVector *qiov,
479 BdrvRequestFlags flags, BlockCompletionFunc *cb, void *opaque);
7b1fb72e 480
88095349
EGE
481 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_flush)(
482 BlockDriverState *bs, BlockCompletionFunc *cb, void *opaque);
9a5a1c62
EGE
483
484 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_pdiscard)(
485 BlockDriverState *bs, int64_t offset, int bytes,
ebc2752b
EGE
486 BlockCompletionFunc *cb, void *opaque);
487
7b1fb72e 488 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_readv)(BlockDriverState *bs,
ebc2752b
EGE
489 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
490
491 /**
492 * @offset: position in bytes to read at
493 * @bytes: number of bytes to read
494 * @qiov: the buffers to fill with read data
495 * @flags: currently unused, always 0
496 *
497 * @offset and @bytes will be a multiple of 'request_alignment',
498 * but the length of individual @qiov elements does not have to
499 * be a multiple.
500 *
501 * @bytes will always equal the total size of @qiov, and will be
502 * no larger than 'max_transfer'.
503 *
504 * The buffer in @qiov may point directly to guest memory.
505 */
7b1fb72e 506 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_preadv)(BlockDriverState *bs,
ebc2752b
EGE
507 int64_t offset, int64_t bytes, QEMUIOVector *qiov,
508 BdrvRequestFlags flags);
509
7b1fb72e
KW
510 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_preadv_part)(
511 BlockDriverState *bs, int64_t offset, int64_t bytes,
ebc2752b
EGE
512 QEMUIOVector *qiov, size_t qiov_offset,
513 BdrvRequestFlags flags);
514
7b1fb72e 515 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_writev)(BlockDriverState *bs,
ebc2752b
EGE
516 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov,
517 int flags);
518 /**
519 * @offset: position in bytes to write at
520 * @bytes: number of bytes to write
521 * @qiov: the buffers containing data to write
522 * @flags: zero or more bits allowed by 'supported_write_flags'
523 *
524 * @offset and @bytes will be a multiple of 'request_alignment',
525 * but the length of individual @qiov elements does not have to
526 * be a multiple.
527 *
528 * @bytes will always equal the total size of @qiov, and will be
529 * no larger than 'max_transfer'.
530 *
531 * The buffer in @qiov may point directly to guest memory.
532 */
7b1fb72e
KW
533 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev)(
534 BlockDriverState *bs, int64_t offset, int64_t bytes, QEMUIOVector *qiov,
ebc2752b 535 BdrvRequestFlags flags);
7b1fb72e
KW
536 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev_part)(
537 BlockDriverState *bs, int64_t offset, int64_t bytes, QEMUIOVector *qiov,
538 size_t qiov_offset, BdrvRequestFlags flags);
ebc2752b
EGE
539
540 /*
541 * Efficiently zero a region of the disk image. Typically an image format
542 * would use a compact metadata representation to implement this. This
543 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
544 * will be called instead.
545 */
abaf8b75
KW
546 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwrite_zeroes)(
547 BlockDriverState *bs, int64_t offset, int64_t bytes,
548 BdrvRequestFlags flags);
9a5a1c62
EGE
549
550 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pdiscard)(
551 BlockDriverState *bs, int64_t offset, int64_t bytes);
ebc2752b
EGE
552
553 /*
554 * Map [offset, offset + nbytes) range onto a child of @bs to copy from,
555 * and invoke bdrv_co_copy_range_from(child, ...), or invoke
556 * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
557 *
558 * See the comment of bdrv_co_copy_range for the parameter and return value
559 * semantics.
560 */
742bf09b
EGE
561 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_copy_range_from)(
562 BlockDriverState *bs, BdrvChild *src, int64_t offset,
563 BdrvChild *dst, int64_t dst_offset, int64_t bytes,
564 BdrvRequestFlags read_flags, BdrvRequestFlags write_flags);
ebc2752b
EGE
565
566 /*
567 * Map [offset, offset + nbytes) range onto a child of bs to copy data to,
568 * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
569 * operation if @bs is the leaf and @src has the same BlockDriver. Return
570 * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
571 *
572 * See the comment of bdrv_co_copy_range for the parameter and return value
573 * semantics.
574 */
742bf09b
EGE
575 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_copy_range_to)(
576 BlockDriverState *bs, BdrvChild *src, int64_t src_offset,
577 BdrvChild *dst, int64_t dst_offset, int64_t bytes,
578 BdrvRequestFlags read_flags, BdrvRequestFlags write_flags);
ebc2752b
EGE
579
580 /*
581 * Building block for bdrv_block_status[_above] and
582 * bdrv_is_allocated[_above]. The driver should answer only
583 * according to the current layer, and should only need to set
584 * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
585 * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
586 * layer, the result should be 0 (and not BDRV_BLOCK_ZERO). See
587 * block.h for the overall meaning of the bits. As a hint, the
588 * flag want_zero is true if the caller cares more about precise
589 * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
590 * overall allocation (favor larger *pnum, perhaps by reporting
591 * _DATA instead of _ZERO). The block layer guarantees input
592 * clamped to bdrv_getlength() and aligned to request_alignment,
593 * as well as non-NULL pnum, map, and file; in turn, the driver
594 * must return an error or set pnum to an aligned non-zero value.
595 *
596 * Note that @bytes is just a hint on how big of a region the
597 * caller wants to inspect. It is not a limit on *pnum.
598 * Implementations are free to return larger values of *pnum if
599 * doing so does not incur a performance penalty.
600 *
601 * block/io.c's bdrv_co_block_status() will utilize an unclamped
602 * *pnum value for the block-status cache on protocol nodes, prior
603 * to clamping *pnum for return to its caller.
604 */
7ff9579e
KW
605 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_block_status)(
606 BlockDriverState *bs,
ebc2752b
EGE
607 bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
608 int64_t *map, BlockDriverState **file);
609
ce14f3b4
VSO
610 /*
611 * Snapshot-access API.
612 *
613 * Block-driver may provide snapshot-access API: special functions to access
614 * some internal "snapshot". The functions are similar with normal
615 * read/block_status/discard handler, but don't have any specific handling
616 * in generic block-layer: no serializing, no alignment, no tracked
617 * requests. So, block-driver that realizes these APIs is fully responsible
618 * for synchronization between snapshot-access API and normal IO requests.
1c14eaab
VSO
619 *
620 * TODO: To be able to support qcow2's internal snapshots, this API will
621 * need to be extended to:
622 * - be able to select a specific snapshot
623 * - receive the snapshot's actual length (which may differ from bs's
624 * length)
ce14f3b4 625 */
7b9e8b22
KW
626 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_preadv_snapshot)(
627 BlockDriverState *bs, int64_t offset, int64_t bytes,
628 QEMUIOVector *qiov, size_t qiov_offset);
629
630 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_snapshot_block_status)(
631 BlockDriverState *bs, bool want_zero, int64_t offset, int64_t bytes,
632 int64_t *pnum, int64_t *map, BlockDriverState **file);
9a5a1c62
EGE
633
634 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pdiscard_snapshot)(
635 BlockDriverState *bs, int64_t offset, int64_t bytes);
ce14f3b4 636
ebc2752b
EGE
637 /*
638 * Invalidate any cached meta-data.
639 */
1b3ff9fe
KW
640 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_invalidate_cache)(
641 BlockDriverState *bs, Error **errp);
ebc2752b
EGE
642
643 /*
644 * Flushes all data for all layers by calling bdrv_co_flush for underlying
645 * layers, if needed. This function is needed for deterministic
646 * synchronization of the flush finishing callback.
647 */
88095349 648 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_flush)(BlockDriverState *bs);
ebc2752b
EGE
649
650 /* Delete a created file. */
48aef794
KW
651 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_delete_file)(
652 BlockDriverState *bs, Error **errp);
ebc2752b
EGE
653
654 /*
655 * Flushes all data that was already written to the OS all the way down to
656 * the disk (for example file-posix.c calls fsync()).
657 */
88095349
EGE
658 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_flush_to_disk)(
659 BlockDriverState *bs);
ebc2752b
EGE
660
661 /*
662 * Flushes all internal caches to the OS. The data may still sit in a
663 * writeback cache of the host OS, but it will survive a crash of the qemu
664 * process.
665 */
88095349
EGE
666 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_flush_to_os)(
667 BlockDriverState *bs);
ebc2752b 668
ebc2752b
EGE
669 /*
670 * Truncate @bs to @offset bytes using the given @prealloc mode
671 * when growing. Modes other than PREALLOC_MODE_OFF should be
672 * rejected when shrinking @bs.
673 *
674 * If @exact is true, @bs must be resized to exactly @offset.
675 * Otherwise, it is sufficient for @bs (if it is a host block
676 * device and thus there is no way to resize it) to be at least
677 * @offset bytes in length.
678 *
679 * If @exact is true and this function fails but would succeed
680 * with @exact = false, it should return -ENOTSUP.
681 */
c2b8e315
KW
682 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_truncate)(
683 BlockDriverState *bs, int64_t offset, bool exact,
684 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
685
8ab8140a
KW
686 int64_t coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_getlength)(
687 BlockDriverState *bs);
688
de335638 689 int64_t coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_get_allocated_file_size)(
82618d7b
EGE
690 BlockDriverState *bs);
691
ebc2752b
EGE
692 BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
693 Error **errp);
694
7b1fb72e
KW
695 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev_compressed)(
696 BlockDriverState *bs, int64_t offset, int64_t bytes,
697 QEMUIOVector *qiov);
698
699 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev_compressed_part)(
700 BlockDriverState *bs, int64_t offset, int64_t bytes,
701 QEMUIOVector *qiov, size_t qiov_offset);
ebc2752b 702
a00e70c0
EGE
703 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_get_info)(
704 BlockDriverState *bs, BlockDriverInfo *bdi);
ebc2752b
EGE
705
706 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
707 Error **errp);
708 BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs);
709
ca5e2ad9 710 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_save_vmstate)(
1b3ff9fe
KW
711 BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
712
ca5e2ad9 713 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_load_vmstate)(
1b3ff9fe 714 BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
ebc2752b 715
ebc2752b 716 /* removable device specific */
c73ff92c
EGE
717 bool coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_is_inserted)(
718 BlockDriverState *bs);
79a292e5
KW
719 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_eject)(
720 BlockDriverState *bs, bool eject_flag);
721 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_lock_medium)(
722 BlockDriverState *bs, bool locked);
ebc2752b
EGE
723
724 /* to control generic scsi devices */
26c518ab
KW
725 BlockAIOCB *coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_aio_ioctl)(
726 BlockDriverState *bs, unsigned long int req, void *buf,
ebc2752b 727 BlockCompletionFunc *cb, void *opaque);
26c518ab
KW
728
729 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_ioctl)(
730 BlockDriverState *bs, unsigned long int req, void *buf);
ebc2752b 731
ebc2752b
EGE
732 /*
733 * Returns 0 for completed check, -errno for internal errors.
734 * The check results are stored in result.
735 */
1b3ff9fe
KW
736 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_check)(
737 BlockDriverState *bs, BdrvCheckResult *result, BdrvCheckMode fix);
ebc2752b 738
cb2bfaa4
EGE
739 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_debug_event)(
740 BlockDriverState *bs, BlkdebugEvent event);
ebc2752b 741
ebc2752b 742 /* io queue for linux-aio */
c3827069
KW
743 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_io_plug)(BlockDriverState *bs);
744 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_io_unplug)(
745 BlockDriverState *bs);
ebc2752b 746
ebc2752b 747 /**
5e8ac217 748 * bdrv_drain_begin is called if implemented in the beginning of a
ebc2752b
EGE
749 * drain operation to drain and stop any internal sources of requests in
750 * the driver.
5e8ac217 751 * bdrv_drain_end is called if implemented at the end of the drain.
ebc2752b
EGE
752 *
753 * They should be used by the driver to e.g. manage scheduled I/O
754 * requests, or toggle an internal state. After the end of the drain new
755 * requests will continue normally.
5e8ac217
KW
756 *
757 * Implementations of both functions must not call aio_poll().
ebc2752b 758 */
5e8ac217
KW
759 void (*bdrv_drain_begin)(BlockDriverState *bs);
760 void (*bdrv_drain_end)(BlockDriverState *bs);
ebc2752b 761
ebc2752b 762 bool (*bdrv_supports_persistent_dirty_bitmap)(BlockDriverState *bs);
167f748d
KW
763
764 bool coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_can_store_new_dirty_bitmap)(
c2d76808
AF
765 BlockDriverState *bs, const char *name, uint32_t granularity,
766 Error **errp);
167f748d
KW
767
768 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_remove_persistent_dirty_bitmap)(
c2d76808 769 BlockDriverState *bs, const char *name, Error **errp);
ebc2752b
EGE
770};
771
7b1fb72e 772static inline bool TSA_NO_TSA block_driver_can_compress(BlockDriver *drv)
ebc2752b
EGE
773{
774 return drv->bdrv_co_pwritev_compressed ||
775 drv->bdrv_co_pwritev_compressed_part;
776}
777
778typedef struct BlockLimits {
779 /*
780 * Alignment requirement, in bytes, for offset/length of I/O
781 * requests. Must be a power of 2 less than INT_MAX; defaults to
782 * 1 for drivers with modern byte interfaces, and to 512
783 * otherwise.
784 */
785 uint32_t request_alignment;
786
787 /*
788 * Maximum number of bytes that can be discarded at once. Must be multiple
789 * of pdiscard_alignment, but need not be power of 2. May be 0 if no
790 * inherent 64-bit limit.
791 */
792 int64_t max_pdiscard;
793
794 /*
795 * Optimal alignment for discard requests in bytes. A power of 2
796 * is best but not mandatory. Must be a multiple of
797 * bl.request_alignment, and must be less than max_pdiscard if
798 * that is set. May be 0 if bl.request_alignment is good enough
799 */
800 uint32_t pdiscard_alignment;
801
802 /*
803 * Maximum number of bytes that can zeroized at once. Must be multiple of
804 * pwrite_zeroes_alignment. 0 means no limit.
805 */
806 int64_t max_pwrite_zeroes;
807
808 /*
809 * Optimal alignment for write zeroes requests in bytes. A power
810 * of 2 is best but not mandatory. Must be a multiple of
811 * bl.request_alignment, and must be less than max_pwrite_zeroes
812 * if that is set. May be 0 if bl.request_alignment is good
813 * enough
814 */
815 uint32_t pwrite_zeroes_alignment;
816
817 /*
818 * Optimal transfer length in bytes. A power of 2 is best but not
819 * mandatory. Must be a multiple of bl.request_alignment, or 0 if
820 * no preferred size
821 */
822 uint32_t opt_transfer;
823
824 /*
825 * Maximal transfer length in bytes. Need not be power of 2, but
826 * must be multiple of opt_transfer and bl.request_alignment, or 0
827 * for no 32-bit limit. For now, anything larger than INT_MAX is
828 * clamped down.
829 */
830 uint32_t max_transfer;
831
832 /*
833 * Maximal hardware transfer length in bytes. Applies whenever
834 * transfers to the device bypass the kernel I/O scheduler, for
835 * example with SG_IO. If larger than max_transfer or if zero,
836 * blk_get_max_hw_transfer will fall back to max_transfer.
837 */
838 uint64_t max_hw_transfer;
839
840 /*
841 * Maximal number of scatter/gather elements allowed by the hardware.
842 * Applies whenever transfers to the device bypass the kernel I/O
843 * scheduler, for example with SG_IO. If larger than max_iov
844 * or if zero, blk_get_max_hw_iov will fall back to max_iov.
845 */
846 int max_hw_iov;
847
848
849 /* memory alignment, in bytes so that no bounce buffer is needed */
850 size_t min_mem_alignment;
851
852 /* memory alignment, in bytes, for bounce buffer */
853 size_t opt_mem_alignment;
854
855 /* maximum number of iovec elements */
856 int max_iov;
160a29e2
PB
857
858 /*
859 * true if the length of the underlying file can change, and QEMU
860 * is expected to adjust automatically. Mostly for CD-ROM drives,
861 * whose length is zero when the tray is empty (they don't need
862 * an explicit monitor command to load the disk inside the guest).
863 */
864 bool has_variable_length;
ebc2752b
EGE
865} BlockLimits;
866
867typedef struct BdrvOpBlocker BdrvOpBlocker;
868
869typedef struct BdrvAioNotifier {
870 void (*attached_aio_context)(AioContext *new_context, void *opaque);
871 void (*detach_aio_context)(void *opaque);
872
873 void *opaque;
874 bool deleted;
875
876 QLIST_ENTRY(BdrvAioNotifier) list;
877} BdrvAioNotifier;
878
879struct BdrvChildClass {
880 /*
881 * If true, bdrv_replace_node() doesn't change the node this BdrvChild
882 * points to.
883 */
884 bool stay_at_node;
885
886 /*
887 * If true, the parent is a BlockDriverState and bdrv_next_all_states()
888 * will return it. This information is used for drain_all, where every node
889 * will be drained separately, so the drain only needs to be propagated to
890 * non-BDS parents.
891 */
892 bool parent_is_bds;
893
abc5a79c
EGE
894 /*
895 * Global state (GS) API. These functions run under the BQL.
896 *
897 * See include/block/block-global-state.h for more information about
898 * the GS API.
899 */
ebc2752b
EGE
900 void (*inherit_options)(BdrvChildRole role, bool parent_is_format,
901 int *child_flags, QDict *child_options,
902 int parent_flags, QDict *parent_options);
ebc2752b 903 void (*change_media)(BdrvChild *child, bool load);
ebc2752b
EGE
904
905 /*
906 * Returns a malloced string that describes the parent of the child for a
907 * human reader. This could be a node-name, BlockBackend name, qdev ID or
908 * QOM path of the device owning the BlockBackend, job type and ID etc. The
909 * caller is responsible for freeing the memory.
910 */
911 char *(*get_parent_desc)(BdrvChild *child);
912
abc5a79c
EGE
913 /*
914 * Notifies the parent that the child has been activated/inactivated (e.g.
915 * when migration is completing) and it can start/stop requesting
916 * permissions and doing I/O on it.
917 */
918 void (*activate)(BdrvChild *child, Error **errp);
919 int (*inactivate)(BdrvChild *child);
920
303de47b
KW
921 void GRAPH_WRLOCK_PTR (*attach)(BdrvChild *child);
922 void GRAPH_WRLOCK_PTR (*detach)(BdrvChild *child);
abc5a79c
EGE
923
924 /*
925 * Notifies the parent that the filename of its child has changed (e.g.
926 * because the direct child was removed from the backing chain), so that it
927 * can update its reference.
928 */
929 int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
930 const char *filename, Error **errp);
931
7e8c182f 932 bool (*change_aio_ctx)(BdrvChild *child, AioContext *ctx,
e08cc001
EGE
933 GHashTable *visited, Transaction *tran,
934 Error **errp);
abc5a79c 935
abc5a79c
EGE
936 /*
937 * I/O API functions. These functions are thread-safe.
938 *
939 * See include/block/block-io.h for more information about
940 * the I/O API.
941 */
942
943 void (*resize)(BdrvChild *child);
944
945 /*
946 * Returns a name that is supposedly more useful for human users than the
947 * node name for identifying the node in question (in particular, a BB
948 * name), or NULL if the parent can't provide a better name.
949 */
950 const char *(*get_name)(BdrvChild *child);
951
d5f8d79c
HR
952 AioContext *(*get_parent_aio_context)(BdrvChild *child);
953
ebc2752b
EGE
954 /*
955 * If this pair of functions is implemented, the parent doesn't issue new
956 * requests after returning from .drained_begin() until .drained_end() is
957 * called.
958 *
959 * These functions must not change the graph (and therefore also must not
960 * call aio_poll(), which could change the graph indirectly).
961 *
ebc2752b
EGE
962 * Note that this can be nested. If drained_begin() was called twice, new
963 * I/O is allowed only after drained_end() was called twice, too.
964 */
965 void (*drained_begin)(BdrvChild *child);
2f65df6e 966 void (*drained_end)(BdrvChild *child);
ebc2752b
EGE
967
968 /*
969 * Returns whether the parent has pending requests for the child. This
970 * callback is polled after .drained_begin() has been called until all
971 * activity on the child has stopped.
972 */
973 bool (*drained_poll)(BdrvChild *child);
ebc2752b
EGE
974};
975
976extern const BdrvChildClass child_of_bds;
977
978struct BdrvChild {
979 BlockDriverState *bs;
980 char *name;
981 const BdrvChildClass *klass;
982 BdrvChildRole role;
983 void *opaque;
984
985 /**
986 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
987 */
988 uint64_t perm;
989
990 /**
991 * Permissions that can still be granted to other users of @bs while this
992 * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
993 */
994 uint64_t shared_perm;
995
996 /*
997 * This link is frozen: the child can neither be replaced nor
998 * detached from the parent.
999 */
1000 bool frozen;
1001
1002 /*
57e05be3 1003 * True if the parent of this child has been drained by this BdrvChild
ebc2752b 1004 * (through klass->drained_*).
57e05be3
KW
1005 *
1006 * It is generally true if bs->quiesce_counter > 0. It may differ while the
ebc2752b
EGE
1007 * child is entering or leaving a drained section.
1008 */
57e05be3 1009 bool quiesced_parent;
ebc2752b
EGE
1010
1011 QLIST_ENTRY(BdrvChild) next;
1012 QLIST_ENTRY(BdrvChild) next_parent;
1013};
1014
1015/*
1016 * Allows bdrv_co_block_status() to cache one data region for a
1017 * protocol node.
1018 *
1019 * @valid: Whether the cache is valid (should be accessed with atomic
1020 * functions so this can be reset by RCU readers)
1021 * @data_start: Offset where we know (or strongly assume) is data
1022 * @data_end: Offset where the data region ends (which is not necessarily
1023 * the start of a zeroed region)
1024 */
1025typedef struct BdrvBlockStatusCache {
1026 struct rcu_head rcu;
1027
1028 bool valid;
1029 int64_t data_start;
1030 int64_t data_end;
1031} BdrvBlockStatusCache;
1032
1033struct BlockDriverState {
1034 /*
1035 * Protected by big QEMU lock or read-only after opening. No special
1036 * locking needed during I/O...
1037 */
1038 int open_flags; /* flags used to open the file, re-used for re-open */
1039 bool encrypted; /* if true, the media is encrypted */
1040 bool sg; /* if true, the device is a /dev/sg* */
1041 bool probed; /* if true, format was probed rather than specified */
1042 bool force_share; /* if true, always allow all shared permissions */
1043 bool implicit; /* if true, this filter node was automatically inserted */
1044
1045 BlockDriver *drv; /* NULL means no media */
1046 void *opaque;
1047
1048 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
1049 /*
1050 * long-running tasks intended to always use the same AioContext as this
1051 * BDS may register themselves in this list to be notified of changes
1052 * regarding this BDS's context
1053 */
1054 QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
1055 bool walking_aio_notifiers; /* to make removal during iteration safe */
1056
1057 char filename[PATH_MAX];
1058 /*
1059 * If not empty, this image is a diff in relation to backing_file.
1060 * Note that this is the name given in the image header and
1061 * therefore may or may not be equal to .backing->bs->filename.
1062 * If this field contains a relative path, it is to be resolved
1063 * relatively to the overlay's location.
1064 */
1065 char backing_file[PATH_MAX];
1066 /*
1067 * The backing filename indicated by the image header. Contrary
1068 * to backing_file, if we ever open this file, auto_backing_file
1069 * is replaced by the resulting BDS's filename (i.e. after a
1070 * bdrv_refresh_filename() run).
1071 */
1072 char auto_backing_file[PATH_MAX];
1073 char backing_format[16]; /* if non-zero and backing_file exists */
1074
1075 QDict *full_open_options;
1076 char exact_filename[PATH_MAX];
1077
ebc2752b
EGE
1078 /* I/O Limits */
1079 BlockLimits bl;
1080
1081 /*
1082 * Flags honored during pread
1083 */
98b3ddc7 1084 BdrvRequestFlags supported_read_flags;
ebc2752b
EGE
1085 /*
1086 * Flags honored during pwrite (so far: BDRV_REQ_FUA,
1087 * BDRV_REQ_WRITE_UNCHANGED).
1088 * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
1089 * writes will be issued as normal writes without the flag set.
1090 * This is important to note for drivers that do not explicitly
1091 * request a WRITE permission for their children and instead take
1092 * the same permissions as their parent did (this is commonly what
1093 * block filters do). Such drivers have to be aware that the
1094 * parent may have taken a WRITE_UNCHANGED permission only and is
1095 * issuing such requests. Drivers either must make sure that
1096 * these requests do not result in plain WRITE accesses (usually
1097 * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
1098 * every incoming write request as-is, including potentially that
1099 * flag), or they have to explicitly take the WRITE permission for
1100 * their children.
1101 */
98b3ddc7 1102 BdrvRequestFlags supported_write_flags;
ebc2752b
EGE
1103 /*
1104 * Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
1105 * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED)
1106 */
98b3ddc7 1107 BdrvRequestFlags supported_zero_flags;
ebc2752b
EGE
1108 /*
1109 * Flags honoured during truncate (so far: BDRV_REQ_ZERO_WRITE).
1110 *
1111 * If BDRV_REQ_ZERO_WRITE is given, the truncate operation must make sure
1112 * that any added space reads as all zeros. If this can't be guaranteed,
1113 * the operation must fail.
1114 */
98b3ddc7 1115 BdrvRequestFlags supported_truncate_flags;
ebc2752b
EGE
1116
1117 /* the following member gives a name to every node on the bs graph. */
1118 char node_name[32];
1119 /* element of the list of named nodes building the graph */
1120 QTAILQ_ENTRY(BlockDriverState) node_list;
1121 /* element of the list of all BlockDriverStates (all_bdrv_states) */
1122 QTAILQ_ENTRY(BlockDriverState) bs_list;
1123 /* element of the list of monitor-owned BDS */
1124 QTAILQ_ENTRY(BlockDriverState) monitor_list;
1125 int refcnt;
1126
1127 /* operation blockers. Protected by BQL. */
1128 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
1129
1130 /*
1131 * The node that this node inherited default options from (and a reopen on
1132 * which can affect this node by changing these defaults). This is always a
1133 * parent node of this node.
1134 */
1135 BlockDriverState *inherits_from;
5bb04747
VSO
1136
1137 /*
1138 * @backing and @file are some of @children or NULL. All these three fields
1139 * (@file, @backing and @children) are modified only in
1140 * bdrv_child_cb_attach() and bdrv_child_cb_detach().
1141 *
1142 * See also comment in include/block/block.h, to learn how backing and file
1143 * are connected with BdrvChildRole.
1144 */
ebc2752b 1145 QLIST_HEAD(, BdrvChild) children;
5bb04747
VSO
1146 BdrvChild *backing;
1147 BdrvChild *file;
1148
ebc2752b
EGE
1149 QLIST_HEAD(, BdrvChild) parents;
1150
1151 QDict *options;
1152 QDict *explicit_options;
1153 BlockdevDetectZeroesOptions detect_zeroes;
1154
1155 /* The error object in use for blocking operations on backing_hd */
1156 Error *backing_blocker;
1157
1158 /* Protected by AioContext lock */
1159
1160 /*
1161 * If we are reading a disk image, give its size in sectors.
1162 * Generally read-only; it is written to by load_snapshot and
1163 * save_snaphost, but the block layer is quiescent during those.
1164 */
1165 int64_t total_sectors;
1166
1167 /* threshold limit for writes, in bytes. "High water mark". */
1168 uint64_t write_threshold_offset;
1169
1170 /*
1171 * Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
1172 * Reading from the list can be done with either the BQL or the
1173 * dirty_bitmap_mutex. Modifying a bitmap only requires
1174 * dirty_bitmap_mutex.
1175 */
1176 QemuMutex dirty_bitmap_mutex;
1177 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
1178
1179 /* Offset after the highest byte written to */
1180 Stat64 wr_highest_offset;
1181
1182 /*
1183 * If true, copy read backing sectors into image. Can be >1 if more
1184 * than one client has requested copy-on-read. Accessed with atomic
1185 * ops.
1186 */
1187 int copy_on_read;
1188
1189 /*
1190 * number of in-flight requests; overall and serialising.
1191 * Accessed with atomic ops.
1192 */
1193 unsigned int in_flight;
1194 unsigned int serialising_in_flight;
1195
1196 /*
1197 * counter for nested bdrv_io_plug.
1198 * Accessed with atomic ops.
1199 */
1200 unsigned io_plugged;
1201
1202 /* do we need to tell the quest if we have a volatile write cache? */
1203 int enable_write_cache;
1204
1205 /* Accessed with atomic ops. */
1206 int quiesce_counter;
ebc2752b
EGE
1207
1208 unsigned int write_gen; /* Current data generation */
1209
1210 /* Protected by reqs_lock. */
1211 CoMutex reqs_lock;
1212 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
1213 CoQueue flush_queue; /* Serializing flush queue */
1214 bool active_flush_req; /* Flush request in flight? */
1215
1216 /* Only read/written by whoever has set active_flush_req to true. */
1217 unsigned int flushed_gen; /* Flushed write generation */
1218
1219 /* BdrvChild links to this node may never be frozen */
1220 bool never_freeze;
1221
1222 /* Lock for block-status cache RCU writers */
1223 CoMutex bsc_modify_lock;
1224 /* Always non-NULL, but must only be dereferenced under an RCU read guard */
1225 BdrvBlockStatusCache *block_status_cache;
1226};
1227
1228struct BlockBackendRootState {
1229 int open_flags;
1230 BlockdevDetectZeroesOptions detect_zeroes;
1231};
1232
1233typedef enum BlockMirrorBackingMode {
1234 /*
1235 * Reuse the existing backing chain from the source for the target.
1236 * - sync=full: Set backing BDS to NULL.
1237 * - sync=top: Use source's backing BDS.
1238 * - sync=none: Use source as the backing BDS.
1239 */
1240 MIRROR_SOURCE_BACKING_CHAIN,
1241
1242 /* Open the target's backing chain completely anew */
1243 MIRROR_OPEN_BACKING_CHAIN,
1244
1245 /* Do not change the target's backing BDS after job completion */
1246 MIRROR_LEAVE_BACKING_CHAIN,
1247} BlockMirrorBackingMode;
1248
1249
1250/*
1251 * Essential block drivers which must always be statically linked into qemu, and
1252 * which therefore can be accessed without using bdrv_find_format()
1253 */
1254extern BlockDriver bdrv_file;
1255extern BlockDriver bdrv_raw;
1256extern BlockDriver bdrv_qcow2;
1257
1258extern unsigned int bdrv_drain_all_count;
1259extern QemuOptsList bdrv_create_opts_simple;
1260
1261/*
1262 * Common functions that are neither I/O nor Global State.
1263 *
04ae220d 1264 * See include/block/block-common.h for more information about
ebc2752b
EGE
1265 * the Common API.
1266 */
1267
1268static inline BlockDriverState *child_bs(BdrvChild *child)
1269{
1270 return child ? child->bs : NULL;
1271}
1272
1273int bdrv_check_request(int64_t offset, int64_t bytes, Error **errp);
69fbfff9 1274char *create_tmp_file(Error **errp);
ebc2752b
EGE
1275void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
1276 QDict *options);
1277
1278
1279int bdrv_check_qiov_request(int64_t offset, int64_t bytes,
1280 QEMUIOVector *qiov, size_t qiov_offset,
1281 Error **errp);
1282
1283#ifdef _WIN32
1284int is_windows_drive(const char *filename);
1285#endif
1286
1287#endif /* BLOCK_INT_COMMON_H */