Merge tag 'for-upstream' of https://repo.or.cz/qemu/kevin into staging
[qemu/ar7.git] / include / block / block-common.h
blob297704c1e966f9776558a4d95c4f459e45e2a0ba
1 /*
2 * QEMU System Emulator block driver
4 * Copyright (c) 2003 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
24 #ifndef BLOCK_COMMON_H
25 #define BLOCK_COMMON_H
27 #include "block/aio.h"
28 #include "block/aio-wait.h"
29 #include "qemu/iov.h"
30 #include "qemu/coroutine.h"
31 #include "block/accounting.h"
32 #include "block/dirty-bitmap.h"
33 #include "block/blockjob.h"
34 #include "qemu/hbitmap.h"
35 #include "qemu/transactions.h"
38 * generated_co_wrapper
40 * Function specifier, which does nothing but mark functions to be
41 * generated by scripts/block-coroutine-wrapper.py
43 * Read more in docs/devel/block-coroutine-wrapper.rst
45 #define generated_co_wrapper
47 /* block.c */
48 typedef struct BlockDriver BlockDriver;
49 typedef struct BdrvChild BdrvChild;
50 typedef struct BdrvChildClass BdrvChildClass;
52 typedef struct BlockDriverInfo {
53 /* in bytes, 0 if irrelevant */
54 int cluster_size;
55 /* offset at which the VM state can be saved (0 if not possible) */
56 int64_t vm_state_offset;
57 bool is_dirty;
59 * True if this block driver only supports compressed writes
61 bool needs_compressed_writes;
62 } BlockDriverInfo;
64 typedef struct BlockFragInfo {
65 uint64_t allocated_clusters;
66 uint64_t total_clusters;
67 uint64_t fragmented_clusters;
68 uint64_t compressed_clusters;
69 } BlockFragInfo;
71 typedef enum {
72 BDRV_REQ_COPY_ON_READ = 0x1,
73 BDRV_REQ_ZERO_WRITE = 0x2,
76 * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
77 * that the block driver should unmap (discard) blocks if it is guaranteed
78 * that the result will read back as zeroes. The flag is only passed to the
79 * driver if the block device is opened with BDRV_O_UNMAP.
81 BDRV_REQ_MAY_UNMAP = 0x4,
84 * An optimization hint when all QEMUIOVector elements are within
85 * previously registered bdrv_register_buf() memory ranges.
87 * Code that replaces the user's QEMUIOVector elements with bounce buffers
88 * must take care to clear this flag.
90 BDRV_REQ_REGISTERED_BUF = 0x8,
92 BDRV_REQ_FUA = 0x10,
93 BDRV_REQ_WRITE_COMPRESSED = 0x20,
96 * Signifies that this write request will not change the visible disk
97 * content.
99 BDRV_REQ_WRITE_UNCHANGED = 0x40,
102 * Forces request serialisation. Use only with write requests.
104 BDRV_REQ_SERIALISING = 0x80,
107 * Execute the request only if the operation can be offloaded or otherwise
108 * be executed efficiently, but return an error instead of using a slow
109 * fallback.
111 BDRV_REQ_NO_FALLBACK = 0x100,
114 * BDRV_REQ_PREFETCH makes sense only in the context of copy-on-read
115 * (i.e., together with the BDRV_REQ_COPY_ON_READ flag or when a COR
116 * filter is involved), in which case it signals that the COR operation
117 * need not read the data into memory (qiov) but only ensure they are
118 * copied to the top layer (i.e., that COR operation is done).
120 BDRV_REQ_PREFETCH = 0x200,
123 * If we need to wait for other requests, just fail immediately. Used
124 * only together with BDRV_REQ_SERIALISING. Used only with requests aligned
125 * to request_alignment (corresponding assertions are in block/io.c).
127 BDRV_REQ_NO_WAIT = 0x400,
129 /* Mask of valid flags */
130 BDRV_REQ_MASK = 0x7ff,
131 } BdrvRequestFlags;
133 #define BDRV_O_NO_SHARE 0x0001 /* don't share permissions */
134 #define BDRV_O_RDWR 0x0002
135 #define BDRV_O_RESIZE 0x0004 /* request permission for resizing the node */
136 #define BDRV_O_SNAPSHOT 0x0008 /* open the file read only and save
137 writes in a snapshot */
138 #define BDRV_O_TEMPORARY 0x0010 /* delete the file after use */
139 #define BDRV_O_NOCACHE 0x0020 /* do not use the host page cache */
140 #define BDRV_O_NATIVE_AIO 0x0080 /* use native AIO instead of the
141 thread pool */
142 #define BDRV_O_NO_BACKING 0x0100 /* don't open the backing file */
143 #define BDRV_O_NO_FLUSH 0x0200 /* disable flushing on this disk */
144 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
145 #define BDRV_O_INACTIVE 0x0800 /* consistency hint for migration handoff */
146 #define BDRV_O_CHECK 0x1000 /* open solely for consistency check */
147 #define BDRV_O_ALLOW_RDWR 0x2000 /* allow reopen to change from r/o to r/w */
148 #define BDRV_O_UNMAP 0x4000 /* execute guest UNMAP/TRIM operations */
149 #define BDRV_O_PROTOCOL 0x8000 /* if no block driver is explicitly given:
150 select an appropriate protocol driver,
151 ignoring the format layer */
152 #define BDRV_O_NO_IO 0x10000 /* don't initialize for I/O */
153 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening
154 read-write fails */
155 #define BDRV_O_IO_URING 0x40000 /* use io_uring instead of the thread pool */
157 #define BDRV_O_CACHE_MASK (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
160 /* Option names of options parsed by the block layer */
162 #define BDRV_OPT_CACHE_WB "cache.writeback"
163 #define BDRV_OPT_CACHE_DIRECT "cache.direct"
164 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
165 #define BDRV_OPT_READ_ONLY "read-only"
166 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
167 #define BDRV_OPT_DISCARD "discard"
168 #define BDRV_OPT_FORCE_SHARE "force-share"
171 #define BDRV_SECTOR_BITS 9
172 #define BDRV_SECTOR_SIZE (1ULL << BDRV_SECTOR_BITS)
174 #define BDRV_REQUEST_MAX_SECTORS MIN_CONST(SIZE_MAX >> BDRV_SECTOR_BITS, \
175 INT_MAX >> BDRV_SECTOR_BITS)
176 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
179 * We want allow aligning requests and disk length up to any 32bit alignment
180 * and don't afraid of overflow.
181 * To achieve it, and in the same time use some pretty number as maximum disk
182 * size, let's define maximum "length" (a limit for any offset/bytes request and
183 * for disk size) to be the greatest power of 2 less than INT64_MAX.
185 #define BDRV_MAX_ALIGNMENT (1L << 30)
186 #define BDRV_MAX_LENGTH (QEMU_ALIGN_DOWN(INT64_MAX, BDRV_MAX_ALIGNMENT))
189 * Allocation status flags for bdrv_block_status() and friends.
191 * Public flags:
192 * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
193 * BDRV_BLOCK_ZERO: offset reads as zero
194 * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
195 * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
196 * layer rather than any backing, set by block layer
197 * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
198 * layer, set by block layer
200 * Internal flags:
201 * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
202 * that the block layer recompute the answer from the returned
203 * BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
204 * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
205 * zeroes in file child of current block node inside
206 * returned region. Only valid together with both
207 * BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
208 * appear with BDRV_BLOCK_ZERO.
210 * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
211 * host offset within the returned BDS that is allocated for the
212 * corresponding raw guest data. However, whether that offset
213 * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
215 * DATA ZERO OFFSET_VALID
216 * t t t sectors read as zero, returned file is zero at offset
217 * t f t sectors read as valid from file at offset
218 * f t t sectors preallocated, read as zero, returned file not
219 * necessarily zero at offset
220 * f f t sectors preallocated but read from backing_hd,
221 * returned file contains garbage at offset
222 * t t f sectors preallocated, read as zero, unknown offset
223 * t f f sectors read from unknown file or offset
224 * f t f not allocated or unknown offset, read as zero
225 * f f f not allocated or unknown offset, read from backing_hd
227 #define BDRV_BLOCK_DATA 0x01
228 #define BDRV_BLOCK_ZERO 0x02
229 #define BDRV_BLOCK_OFFSET_VALID 0x04
230 #define BDRV_BLOCK_RAW 0x08
231 #define BDRV_BLOCK_ALLOCATED 0x10
232 #define BDRV_BLOCK_EOF 0x20
233 #define BDRV_BLOCK_RECURSE 0x40
235 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
237 typedef struct BDRVReopenState {
238 BlockDriverState *bs;
239 int flags;
240 BlockdevDetectZeroesOptions detect_zeroes;
241 bool backing_missing;
242 BlockDriverState *old_backing_bs; /* keep pointer for permissions update */
243 BlockDriverState *old_file_bs; /* keep pointer for permissions update */
244 QDict *options;
245 QDict *explicit_options;
246 void *opaque;
247 } BDRVReopenState;
250 * Block operation types
252 typedef enum BlockOpType {
253 BLOCK_OP_TYPE_BACKUP_SOURCE,
254 BLOCK_OP_TYPE_BACKUP_TARGET,
255 BLOCK_OP_TYPE_CHANGE,
256 BLOCK_OP_TYPE_COMMIT_SOURCE,
257 BLOCK_OP_TYPE_COMMIT_TARGET,
258 BLOCK_OP_TYPE_DATAPLANE,
259 BLOCK_OP_TYPE_DRIVE_DEL,
260 BLOCK_OP_TYPE_EJECT,
261 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
262 BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
263 BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
264 BLOCK_OP_TYPE_MIRROR_SOURCE,
265 BLOCK_OP_TYPE_MIRROR_TARGET,
266 BLOCK_OP_TYPE_RESIZE,
267 BLOCK_OP_TYPE_STREAM,
268 BLOCK_OP_TYPE_REPLACE,
269 BLOCK_OP_TYPE_MAX,
270 } BlockOpType;
272 /* Block node permission constants */
273 enum {
275 * A user that has the "permission" of consistent reads is guaranteed that
276 * their view of the contents of the block device is complete and
277 * self-consistent, representing the contents of a disk at a specific
278 * point.
280 * For most block devices (including their backing files) this is true, but
281 * the property cannot be maintained in a few situations like for
282 * intermediate nodes of a commit block job.
284 BLK_PERM_CONSISTENT_READ = 0x01,
286 /** This permission is required to change the visible disk contents. */
287 BLK_PERM_WRITE = 0x02,
290 * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
291 * required for writes to the block node when the caller promises that
292 * the visible disk content doesn't change.
294 * As the BLK_PERM_WRITE permission is strictly stronger, either is
295 * sufficient to perform an unchanging write.
297 BLK_PERM_WRITE_UNCHANGED = 0x04,
299 /** This permission is required to change the size of a block node. */
300 BLK_PERM_RESIZE = 0x08,
303 * There was a now-removed bit BLK_PERM_GRAPH_MOD, with value of 0x10. QEMU
304 * 6.1 and earlier may still lock the corresponding byte in block/file-posix
305 * locking. So, implementing some new permission should be very careful to
306 * not interfere with this old unused thing.
309 BLK_PERM_ALL = 0x0f,
311 DEFAULT_PERM_PASSTHROUGH = BLK_PERM_CONSISTENT_READ
312 | BLK_PERM_WRITE
313 | BLK_PERM_WRITE_UNCHANGED
314 | BLK_PERM_RESIZE,
316 DEFAULT_PERM_UNCHANGED = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
320 * Flags that parent nodes assign to child nodes to specify what kind of
321 * role(s) they take.
323 * At least one of DATA, METADATA, FILTERED, or COW must be set for
324 * every child.
327 * = Connection with bs->children, bs->file and bs->backing fields =
329 * 1. Filters
331 * Filter drivers have drv->is_filter = true.
333 * Filter node has exactly one FILTERED|PRIMARY child, and may have other
334 * children which must not have these bits (one example is the
335 * copy-before-write filter, which also has its target DATA child).
337 * Filter nodes never have COW children.
339 * For most filters, the filtered child is linked in bs->file, bs->backing is
340 * NULL. For some filters (as an exception), it is the other way around; those
341 * drivers will have drv->filtered_child_is_backing set to true (see that
342 * field’s documentation for what drivers this concerns)
344 * 2. "raw" driver (block/raw-format.c)
346 * Formally it's not a filter (drv->is_filter = false)
348 * bs->backing is always NULL
350 * Only has one child, linked in bs->file. Its role is either FILTERED|PRIMARY
351 * (like filter) or DATA|PRIMARY depending on options.
353 * 3. Other drivers
355 * Don't have any FILTERED children.
357 * May have at most one COW child. In this case it's linked in bs->backing.
358 * Otherwise bs->backing is NULL. COW child is never PRIMARY.
360 * May have at most one PRIMARY child. In this case it's linked in bs->file.
361 * Otherwise bs->file is NULL.
363 * May also have some other children that don't have the PRIMARY or COW bit set.
365 enum BdrvChildRoleBits {
367 * This child stores data.
368 * Any node may have an arbitrary number of such children.
370 BDRV_CHILD_DATA = (1 << 0),
373 * This child stores metadata.
374 * Any node may have an arbitrary number of metadata-storing
375 * children.
377 BDRV_CHILD_METADATA = (1 << 1),
380 * A child that always presents exactly the same visible data as
381 * the parent, e.g. by virtue of the parent forwarding all reads
382 * and writes.
383 * This flag is mutually exclusive with DATA, METADATA, and COW.
384 * Any node may have at most one filtered child at a time.
386 BDRV_CHILD_FILTERED = (1 << 2),
389 * Child from which to read all data that isn't allocated in the
390 * parent (i.e., the backing child); such data is copied to the
391 * parent through COW (and optionally COR).
392 * This field is mutually exclusive with DATA, METADATA, and
393 * FILTERED.
394 * Any node may have at most one such backing child at a time.
396 BDRV_CHILD_COW = (1 << 3),
399 * The primary child. For most drivers, this is the child whose
400 * filename applies best to the parent node.
401 * Any node may have at most one primary child at a time.
403 BDRV_CHILD_PRIMARY = (1 << 4),
405 /* Useful combination of flags */
406 BDRV_CHILD_IMAGE = BDRV_CHILD_DATA
407 | BDRV_CHILD_METADATA
408 | BDRV_CHILD_PRIMARY,
411 /* Mask of BdrvChildRoleBits values */
412 typedef unsigned int BdrvChildRole;
414 typedef struct BdrvCheckResult {
415 int corruptions;
416 int leaks;
417 int check_errors;
418 int corruptions_fixed;
419 int leaks_fixed;
420 int64_t image_end_offset;
421 BlockFragInfo bfi;
422 } BdrvCheckResult;
424 typedef enum {
425 BDRV_FIX_LEAKS = 1,
426 BDRV_FIX_ERRORS = 2,
427 } BdrvCheckMode;
429 typedef struct BlockSizes {
430 uint32_t phys;
431 uint32_t log;
432 } BlockSizes;
434 typedef struct HDGeometry {
435 uint32_t heads;
436 uint32_t sectors;
437 uint32_t cylinders;
438 } HDGeometry;
441 * Common functions that are neither I/O nor Global State.
443 * These functions must never call any function from other categories
444 * (I/O, "I/O or GS", Global State) except this one, but can be invoked by
445 * all of them.
448 char *bdrv_perm_names(uint64_t perm);
449 uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm);
451 void bdrv_init_with_whitelist(void);
452 bool bdrv_uses_whitelist(void);
453 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
455 int bdrv_parse_aio(const char *mode, int *flags);
456 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
457 int bdrv_parse_discard_flags(const char *mode, int *flags);
459 int path_has_protocol(const char *path);
460 int path_is_absolute(const char *path);
461 char *path_combine(const char *base_path, const char *filename);
463 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
464 const char *backing,
465 Error **errp);
467 #endif /* BLOCK_COMMON_H */