2 * Block driver for the QCOW version 2 format
4 * Copyright (c) 2004-2006 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
28 #include "crypto/block.h"
29 #include "qemu/coroutine.h"
30 #include "qemu/units.h"
31 #include "block/block_int.h"
34 //#define DEBUG_ALLOC2
37 #define QCOW_MAGIC (('Q' << 24) | ('F' << 16) | ('I' << 8) | 0xfb)
39 #define QCOW_CRYPT_NONE 0
40 #define QCOW_CRYPT_AES 1
41 #define QCOW_CRYPT_LUKS 2
43 #define QCOW_MAX_CRYPT_CLUSTERS 32
44 #define QCOW_MAX_SNAPSHOTS 65536
46 /* Field widths in qcow2 mean normal cluster offsets cannot reach
47 * 64PB; depending on cluster size, compressed clusters can have a
48 * smaller limit (64PB for up to 16k clusters, then ramps down to
49 * 512TB for 2M clusters). */
50 #define QCOW_MAX_CLUSTER_OFFSET ((1ULL << 56) - 1)
52 /* 8 MB refcount table is enough for 2 PB images at 64k cluster size
53 * (128 GB for 512 byte clusters, 2 EB for 2 MB clusters) */
54 #define QCOW_MAX_REFTABLE_SIZE (8 * MiB)
56 /* 32 MB L1 table is enough for 2 PB images at 64k cluster size
57 * (128 GB for 512 byte clusters, 2 EB for 2 MB clusters) */
58 #define QCOW_MAX_L1_SIZE (32 * MiB)
60 /* Allow for an average of 1k per snapshot table entry, should be plenty of
61 * space for snapshot names and IDs */
62 #define QCOW_MAX_SNAPSHOTS_SIZE (1024 * QCOW_MAX_SNAPSHOTS)
64 /* Maximum amount of extra data per snapshot table entry to accept */
65 #define QCOW_MAX_SNAPSHOT_EXTRA_DATA 1024
67 /* Bitmap header extension constraints */
68 #define QCOW2_MAX_BITMAPS 65535
69 #define QCOW2_MAX_BITMAP_DIRECTORY_SIZE (1024 * QCOW2_MAX_BITMAPS)
71 /* Maximum of parallel sub-request per guest request */
72 #define QCOW2_MAX_WORKERS 8
74 /* indicate that the refcount of the referenced cluster is exactly one. */
75 #define QCOW_OFLAG_COPIED (1ULL << 63)
76 /* indicate that the cluster is compressed (they never have the copied flag) */
77 #define QCOW_OFLAG_COMPRESSED (1ULL << 62)
78 /* The cluster reads as all zeros */
79 #define QCOW_OFLAG_ZERO (1ULL << 0)
81 #define QCOW_EXTL2_SUBCLUSTERS_PER_CLUSTER 32
83 /* The subcluster X [0..31] is allocated */
84 #define QCOW_OFLAG_SUB_ALLOC(X) (1ULL << (X))
85 /* The subcluster X [0..31] reads as zeroes */
86 #define QCOW_OFLAG_SUB_ZERO(X) (QCOW_OFLAG_SUB_ALLOC(X) << 32)
87 /* Subclusters [X, Y) (0 <= X <= Y <= 32) are allocated */
88 #define QCOW_OFLAG_SUB_ALLOC_RANGE(X, Y) \
89 (QCOW_OFLAG_SUB_ALLOC(Y) - QCOW_OFLAG_SUB_ALLOC(X))
90 /* Subclusters [X, Y) (0 <= X <= Y <= 32) read as zeroes */
91 #define QCOW_OFLAG_SUB_ZERO_RANGE(X, Y) \
92 (QCOW_OFLAG_SUB_ALLOC_RANGE(X, Y) << 32)
93 /* L2 entry bitmap with all allocation bits set */
94 #define QCOW_L2_BITMAP_ALL_ALLOC (QCOW_OFLAG_SUB_ALLOC_RANGE(0, 32))
95 /* L2 entry bitmap with all "read as zeroes" bits set */
96 #define QCOW_L2_BITMAP_ALL_ZEROES (QCOW_OFLAG_SUB_ZERO_RANGE(0, 32))
98 /* Size of normal and extended L2 entries */
99 #define L2E_SIZE_NORMAL (sizeof(uint64_t))
100 #define L2E_SIZE_EXTENDED (sizeof(uint64_t) * 2)
102 /* Size of L1 table entries */
103 #define L1E_SIZE (sizeof(uint64_t))
105 /* Size of reftable entries */
106 #define REFTABLE_ENTRY_SIZE (sizeof(uint64_t))
108 #define MIN_CLUSTER_BITS 9
109 #define MAX_CLUSTER_BITS 21
111 /* Defined in the qcow2 spec (compressed cluster descriptor) */
112 #define QCOW2_COMPRESSED_SECTOR_SIZE 512U
113 #define QCOW2_COMPRESSED_SECTOR_MASK (~(QCOW2_COMPRESSED_SECTOR_SIZE - 1ULL))
115 /* Must be at least 2 to cover COW */
116 #define MIN_L2_CACHE_SIZE 2 /* cache entries */
118 /* Must be at least 4 to cover all cases of refcount table growth */
119 #define MIN_REFCOUNT_CACHE_SIZE 4 /* clusters */
122 #define DEFAULT_L2_CACHE_MAX_SIZE (32 * MiB)
123 #define DEFAULT_CACHE_CLEAN_INTERVAL 600 /* seconds */
125 #define DEFAULT_L2_CACHE_MAX_SIZE (8 * MiB)
126 /* Cache clean interval is currently available only on Linux, so must be 0 */
127 #define DEFAULT_CACHE_CLEAN_INTERVAL 0
130 #define DEFAULT_CLUSTER_SIZE 65536
132 #define QCOW2_OPT_DATA_FILE "data-file"
133 #define QCOW2_OPT_LAZY_REFCOUNTS "lazy-refcounts"
134 #define QCOW2_OPT_DISCARD_REQUEST "pass-discard-request"
135 #define QCOW2_OPT_DISCARD_SNAPSHOT "pass-discard-snapshot"
136 #define QCOW2_OPT_DISCARD_OTHER "pass-discard-other"
137 #define QCOW2_OPT_OVERLAP "overlap-check"
138 #define QCOW2_OPT_OVERLAP_TEMPLATE "overlap-check.template"
139 #define QCOW2_OPT_OVERLAP_MAIN_HEADER "overlap-check.main-header"
140 #define QCOW2_OPT_OVERLAP_ACTIVE_L1 "overlap-check.active-l1"
141 #define QCOW2_OPT_OVERLAP_ACTIVE_L2 "overlap-check.active-l2"
142 #define QCOW2_OPT_OVERLAP_REFCOUNT_TABLE "overlap-check.refcount-table"
143 #define QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK "overlap-check.refcount-block"
144 #define QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE "overlap-check.snapshot-table"
145 #define QCOW2_OPT_OVERLAP_INACTIVE_L1 "overlap-check.inactive-l1"
146 #define QCOW2_OPT_OVERLAP_INACTIVE_L2 "overlap-check.inactive-l2"
147 #define QCOW2_OPT_OVERLAP_BITMAP_DIRECTORY "overlap-check.bitmap-directory"
148 #define QCOW2_OPT_CACHE_SIZE "cache-size"
149 #define QCOW2_OPT_L2_CACHE_SIZE "l2-cache-size"
150 #define QCOW2_OPT_L2_CACHE_ENTRY_SIZE "l2-cache-entry-size"
151 #define QCOW2_OPT_REFCOUNT_CACHE_SIZE "refcount-cache-size"
152 #define QCOW2_OPT_CACHE_CLEAN_INTERVAL "cache-clean-interval"
154 typedef struct QCowHeader
{
157 uint64_t backing_file_offset
;
158 uint32_t backing_file_size
;
159 uint32_t cluster_bits
;
160 uint64_t size
; /* in bytes */
161 uint32_t crypt_method
;
162 uint32_t l1_size
; /* XXX: save number of clusters instead ? */
163 uint64_t l1_table_offset
;
164 uint64_t refcount_table_offset
;
165 uint32_t refcount_table_clusters
;
166 uint32_t nb_snapshots
;
167 uint64_t snapshots_offset
;
169 /* The following fields are only valid for version >= 3 */
170 uint64_t incompatible_features
;
171 uint64_t compatible_features
;
172 uint64_t autoclear_features
;
174 uint32_t refcount_order
;
175 uint32_t header_length
;
177 /* Additional fields */
178 uint8_t compression_type
;
180 /* header must be a multiple of 8 */
182 } QEMU_PACKED QCowHeader
;
184 QEMU_BUILD_BUG_ON(!QEMU_IS_ALIGNED(sizeof(QCowHeader
), 8));
186 typedef struct QEMU_PACKED QCowSnapshotHeader
{
187 /* header is 8 byte aligned */
188 uint64_t l1_table_offset
;
191 uint16_t id_str_size
;
197 uint64_t vm_clock_nsec
;
199 uint32_t vm_state_size
;
200 uint32_t extra_data_size
; /* for extension */
201 /* extra data follows */
204 } QCowSnapshotHeader
;
206 typedef struct QEMU_PACKED QCowSnapshotExtraData
{
207 uint64_t vm_state_size_large
;
210 } QCowSnapshotExtraData
;
213 typedef struct QCowSnapshot
{
214 uint64_t l1_table_offset
;
219 uint64_t vm_state_size
;
222 uint64_t vm_clock_nsec
;
223 /* icount value for the moment when snapshot was taken */
225 /* Size of all extra data, including QCowSnapshotExtraData if available */
226 uint32_t extra_data_size
;
227 /* Data beyond QCowSnapshotExtraData, if any */
228 void *unknown_extra_data
;
232 typedef struct Qcow2Cache Qcow2Cache
;
234 typedef struct Qcow2CryptoHeaderExtension
{
237 } QEMU_PACKED Qcow2CryptoHeaderExtension
;
239 typedef struct Qcow2UnknownHeaderExtension
{
242 QLIST_ENTRY(Qcow2UnknownHeaderExtension
) next
;
244 } Qcow2UnknownHeaderExtension
;
247 QCOW2_FEAT_TYPE_INCOMPATIBLE
= 0,
248 QCOW2_FEAT_TYPE_COMPATIBLE
= 1,
249 QCOW2_FEAT_TYPE_AUTOCLEAR
= 2,
252 /* Incompatible feature bits */
254 QCOW2_INCOMPAT_DIRTY_BITNR
= 0,
255 QCOW2_INCOMPAT_CORRUPT_BITNR
= 1,
256 QCOW2_INCOMPAT_DATA_FILE_BITNR
= 2,
257 QCOW2_INCOMPAT_COMPRESSION_BITNR
= 3,
258 QCOW2_INCOMPAT_EXTL2_BITNR
= 4,
259 QCOW2_INCOMPAT_DIRTY
= 1 << QCOW2_INCOMPAT_DIRTY_BITNR
,
260 QCOW2_INCOMPAT_CORRUPT
= 1 << QCOW2_INCOMPAT_CORRUPT_BITNR
,
261 QCOW2_INCOMPAT_DATA_FILE
= 1 << QCOW2_INCOMPAT_DATA_FILE_BITNR
,
262 QCOW2_INCOMPAT_COMPRESSION
= 1 << QCOW2_INCOMPAT_COMPRESSION_BITNR
,
263 QCOW2_INCOMPAT_EXTL2
= 1 << QCOW2_INCOMPAT_EXTL2_BITNR
,
265 QCOW2_INCOMPAT_MASK
= QCOW2_INCOMPAT_DIRTY
266 | QCOW2_INCOMPAT_CORRUPT
267 | QCOW2_INCOMPAT_DATA_FILE
268 | QCOW2_INCOMPAT_COMPRESSION
269 | QCOW2_INCOMPAT_EXTL2
,
272 /* Compatible feature bits */
274 QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR
= 0,
275 QCOW2_COMPAT_LAZY_REFCOUNTS
= 1 << QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR
,
277 QCOW2_COMPAT_FEAT_MASK
= QCOW2_COMPAT_LAZY_REFCOUNTS
,
280 /* Autoclear feature bits */
282 QCOW2_AUTOCLEAR_BITMAPS_BITNR
= 0,
283 QCOW2_AUTOCLEAR_DATA_FILE_RAW_BITNR
= 1,
284 QCOW2_AUTOCLEAR_BITMAPS
= 1 << QCOW2_AUTOCLEAR_BITMAPS_BITNR
,
285 QCOW2_AUTOCLEAR_DATA_FILE_RAW
= 1 << QCOW2_AUTOCLEAR_DATA_FILE_RAW_BITNR
,
287 QCOW2_AUTOCLEAR_MASK
= QCOW2_AUTOCLEAR_BITMAPS
288 | QCOW2_AUTOCLEAR_DATA_FILE_RAW
,
291 enum qcow2_discard_type
{
292 QCOW2_DISCARD_NEVER
= 0,
293 QCOW2_DISCARD_ALWAYS
,
294 QCOW2_DISCARD_REQUEST
,
295 QCOW2_DISCARD_SNAPSHOT
,
300 typedef struct Qcow2Feature
{
304 } QEMU_PACKED Qcow2Feature
;
306 typedef struct Qcow2DiscardRegion
{
307 BlockDriverState
*bs
;
310 QTAILQ_ENTRY(Qcow2DiscardRegion
) next
;
311 } Qcow2DiscardRegion
;
313 typedef uint64_t Qcow2GetRefcountFunc(const void *refcount_array
,
315 typedef void Qcow2SetRefcountFunc(void *refcount_array
,
316 uint64_t index
, uint64_t value
);
318 typedef struct Qcow2BitmapHeaderExt
{
321 uint64_t bitmap_directory_size
;
322 uint64_t bitmap_directory_offset
;
323 } QEMU_PACKED Qcow2BitmapHeaderExt
;
325 #define QCOW2_MAX_THREADS 4
327 typedef struct BDRVQcow2State
{
333 int subclusters_per_cluster
;
337 int l1_vm_state_index
;
338 int refcount_block_bits
;
339 int refcount_block_size
;
342 uint64_t cluster_offset_mask
;
343 uint64_t l1_table_offset
;
346 Qcow2Cache
*l2_table_cache
;
347 Qcow2Cache
*refcount_block_cache
;
348 QEMUTimer
*cache_clean_timer
;
349 unsigned cache_clean_interval
;
351 QLIST_HEAD(, QCowL2Meta
) cluster_allocs
;
353 uint64_t *refcount_table
;
354 uint64_t refcount_table_offset
;
355 uint32_t refcount_table_size
;
356 uint32_t max_refcount_table_index
; /* Last used entry in refcount_table */
357 uint64_t free_cluster_index
;
358 uint64_t free_byte_offset
;
362 Qcow2CryptoHeaderExtension crypto_header
; /* QCow2 header extension */
363 QCryptoBlockOpenOptions
*crypto_opts
; /* Disk encryption runtime options */
364 QCryptoBlock
*crypto
; /* Disk encryption format driver */
365 bool crypt_physical_offset
; /* Whether to use virtual or physical offset
366 for encryption initialization vector tweak */
367 uint32_t crypt_method_header
;
368 uint64_t snapshots_offset
;
370 unsigned int nb_snapshots
;
371 QCowSnapshot
*snapshots
;
374 uint64_t bitmap_directory_size
;
375 uint64_t bitmap_directory_offset
;
379 bool use_lazy_refcounts
;
382 uint64_t refcount_max
;
384 Qcow2GetRefcountFunc
*get_refcount
;
385 Qcow2SetRefcountFunc
*set_refcount
;
387 bool discard_passthrough
[QCOW2_DISCARD_MAX
];
389 int overlap_check
; /* bitmask of Qcow2MetadataOverlap values */
390 bool signaled_corruption
;
392 uint64_t incompatible_features
;
393 uint64_t compatible_features
;
394 uint64_t autoclear_features
;
396 size_t unknown_header_fields_size
;
397 void *unknown_header_fields
;
398 QLIST_HEAD(, Qcow2UnknownHeaderExtension
) unknown_header_ext
;
399 QTAILQ_HEAD (, Qcow2DiscardRegion
) discards
;
402 /* Backing file path and format as stored in the image (this is not the
403 * effective path/format, which may be the result of a runtime option
405 char *image_backing_file
;
406 char *image_backing_format
;
407 char *image_data_file
;
409 CoQueue thread_task_queue
;
412 BdrvChild
*data_file
;
414 bool metadata_preallocation_checked
;
415 bool metadata_preallocation
;
417 * Compression type used for the image. Default: 0 - ZLIB
418 * The image compression type is set on image creation.
419 * For now, the only way to change the compression type
420 * is to convert the image with the desired compression type set.
422 Qcow2CompressionType compression_type
;
425 typedef struct Qcow2COWRegion
{
427 * Offset of the COW region in bytes from the start of the first cluster
428 * touched by the request.
432 /** Number of bytes to copy */
437 * Describes an in-flight (part of a) write request that writes to clusters
438 * that need to have their L2 table entries updated (because they are
439 * newly allocated or need changes in their L2 bitmaps)
441 typedef struct QCowL2Meta
443 /** Guest offset of the first updated cluster */
446 /** Host offset of the first updated cluster */
447 uint64_t alloc_offset
;
449 /** Number of updated clusters */
452 /** Do not free the old clusters */
453 bool keep_old_clusters
;
456 * Requests that overlap with this allocation and wait to be restarted
457 * when the allocating request has completed.
459 CoQueue dependent_requests
;
462 * The COW Region immediately before the area the guest actually
463 * writes to. This (part of the) write request starts at
464 * cow_start.offset + cow_start.nb_bytes.
466 Qcow2COWRegion cow_start
;
469 * The COW Region immediately after the area the guest actually
470 * writes to. This (part of the) write request ends at cow_end.offset
471 * (which must always be set even when cow_end.nb_bytes is 0).
473 Qcow2COWRegion cow_end
;
476 * Indicates that COW regions are already handled and do not require
477 * any more processing.
482 * Indicates that this is not a normal write request but a preallocation.
483 * If the image has extended L2 entries this means that no new individual
484 * subclusters will be marked as allocated in the L2 bitmap (but any
485 * existing contents of that bitmap will be kept).
490 * The I/O vector with the data from the actual guest write request.
491 * If non-NULL, this is meant to be merged together with the data
492 * from @cow_start and @cow_end into one single write operation.
494 QEMUIOVector
*data_qiov
;
495 size_t data_qiov_offset
;
497 /** Pointer to next L2Meta of the same write request */
498 struct QCowL2Meta
*next
;
500 QLIST_ENTRY(QCowL2Meta
) next_in_flight
;
504 * In images with standard L2 entries all clusters are treated as if
505 * they had one subcluster so QCow2ClusterType and QCow2SubclusterType
506 * can be mapped to each other and have the exact same meaning
507 * (QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC cannot happen in these images).
509 * In images with extended L2 entries QCow2ClusterType refers to the
510 * complete cluster and QCow2SubclusterType to each of the individual
511 * subclusters, so there are several possible combinations:
513 * |--------------+---------------------------|
514 * | Cluster type | Possible subcluster types |
515 * |--------------+---------------------------|
516 * | UNALLOCATED | UNALLOCATED_PLAIN |
518 * |--------------+---------------------------|
519 * | NORMAL | UNALLOCATED_ALLOC |
522 * |--------------+---------------------------|
523 * | COMPRESSED | COMPRESSED |
524 * |--------------+---------------------------|
526 * QCOW2_SUBCLUSTER_INVALID means that the L2 entry is incorrect and
527 * the image should be marked corrupt.
530 typedef enum QCow2ClusterType
{
531 QCOW2_CLUSTER_UNALLOCATED
,
532 QCOW2_CLUSTER_ZERO_PLAIN
,
533 QCOW2_CLUSTER_ZERO_ALLOC
,
534 QCOW2_CLUSTER_NORMAL
,
535 QCOW2_CLUSTER_COMPRESSED
,
538 typedef enum QCow2SubclusterType
{
539 QCOW2_SUBCLUSTER_UNALLOCATED_PLAIN
,
540 QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC
,
541 QCOW2_SUBCLUSTER_ZERO_PLAIN
,
542 QCOW2_SUBCLUSTER_ZERO_ALLOC
,
543 QCOW2_SUBCLUSTER_NORMAL
,
544 QCOW2_SUBCLUSTER_COMPRESSED
,
545 QCOW2_SUBCLUSTER_INVALID
,
546 } QCow2SubclusterType
;
548 typedef enum QCow2MetadataOverlap
{
549 QCOW2_OL_MAIN_HEADER_BITNR
= 0,
550 QCOW2_OL_ACTIVE_L1_BITNR
= 1,
551 QCOW2_OL_ACTIVE_L2_BITNR
= 2,
552 QCOW2_OL_REFCOUNT_TABLE_BITNR
= 3,
553 QCOW2_OL_REFCOUNT_BLOCK_BITNR
= 4,
554 QCOW2_OL_SNAPSHOT_TABLE_BITNR
= 5,
555 QCOW2_OL_INACTIVE_L1_BITNR
= 6,
556 QCOW2_OL_INACTIVE_L2_BITNR
= 7,
557 QCOW2_OL_BITMAP_DIRECTORY_BITNR
= 8,
559 QCOW2_OL_MAX_BITNR
= 9,
562 QCOW2_OL_MAIN_HEADER
= (1 << QCOW2_OL_MAIN_HEADER_BITNR
),
563 QCOW2_OL_ACTIVE_L1
= (1 << QCOW2_OL_ACTIVE_L1_BITNR
),
564 QCOW2_OL_ACTIVE_L2
= (1 << QCOW2_OL_ACTIVE_L2_BITNR
),
565 QCOW2_OL_REFCOUNT_TABLE
= (1 << QCOW2_OL_REFCOUNT_TABLE_BITNR
),
566 QCOW2_OL_REFCOUNT_BLOCK
= (1 << QCOW2_OL_REFCOUNT_BLOCK_BITNR
),
567 QCOW2_OL_SNAPSHOT_TABLE
= (1 << QCOW2_OL_SNAPSHOT_TABLE_BITNR
),
568 QCOW2_OL_INACTIVE_L1
= (1 << QCOW2_OL_INACTIVE_L1_BITNR
),
569 /* NOTE: Checking overlaps with inactive L2 tables will result in bdrv
571 QCOW2_OL_INACTIVE_L2
= (1 << QCOW2_OL_INACTIVE_L2_BITNR
),
572 QCOW2_OL_BITMAP_DIRECTORY
= (1 << QCOW2_OL_BITMAP_DIRECTORY_BITNR
),
573 } QCow2MetadataOverlap
;
575 /* Perform all overlap checks which can be done in constant time */
576 #define QCOW2_OL_CONSTANT \
577 (QCOW2_OL_MAIN_HEADER | QCOW2_OL_ACTIVE_L1 | QCOW2_OL_REFCOUNT_TABLE | \
578 QCOW2_OL_SNAPSHOT_TABLE | QCOW2_OL_BITMAP_DIRECTORY)
580 /* Perform all overlap checks which don't require disk access */
581 #define QCOW2_OL_CACHED \
582 (QCOW2_OL_CONSTANT | QCOW2_OL_ACTIVE_L2 | QCOW2_OL_REFCOUNT_BLOCK | \
583 QCOW2_OL_INACTIVE_L1)
585 /* Perform all overlap checks */
586 #define QCOW2_OL_ALL \
587 (QCOW2_OL_CACHED | QCOW2_OL_INACTIVE_L2)
589 #define L1E_OFFSET_MASK 0x00fffffffffffe00ULL
590 #define L2E_OFFSET_MASK 0x00fffffffffffe00ULL
591 #define L2E_COMPRESSED_OFFSET_SIZE_MASK 0x3fffffffffffffffULL
593 #define REFT_OFFSET_MASK 0xfffffffffffffe00ULL
595 #define INV_OFFSET (-1ULL)
597 static inline bool has_subclusters(BDRVQcow2State
*s
)
599 return s
->incompatible_features
& QCOW2_INCOMPAT_EXTL2
;
602 static inline size_t l2_entry_size(BDRVQcow2State
*s
)
604 return has_subclusters(s
) ? L2E_SIZE_EXTENDED
: L2E_SIZE_NORMAL
;
607 static inline uint64_t get_l2_entry(BDRVQcow2State
*s
, uint64_t *l2_slice
,
610 idx
*= l2_entry_size(s
) / sizeof(uint64_t);
611 return be64_to_cpu(l2_slice
[idx
]);
614 static inline uint64_t get_l2_bitmap(BDRVQcow2State
*s
, uint64_t *l2_slice
,
617 if (has_subclusters(s
)) {
618 idx
*= l2_entry_size(s
) / sizeof(uint64_t);
619 return be64_to_cpu(l2_slice
[idx
+ 1]);
621 return 0; /* For convenience only; this value has no meaning. */
625 static inline void set_l2_entry(BDRVQcow2State
*s
, uint64_t *l2_slice
,
626 int idx
, uint64_t entry
)
628 idx
*= l2_entry_size(s
) / sizeof(uint64_t);
629 l2_slice
[idx
] = cpu_to_be64(entry
);
632 static inline void set_l2_bitmap(BDRVQcow2State
*s
, uint64_t *l2_slice
,
633 int idx
, uint64_t bitmap
)
635 assert(has_subclusters(s
));
636 idx
*= l2_entry_size(s
) / sizeof(uint64_t);
637 l2_slice
[idx
+ 1] = cpu_to_be64(bitmap
);
640 static inline bool has_data_file(BlockDriverState
*bs
)
642 BDRVQcow2State
*s
= bs
->opaque
;
643 return (s
->data_file
!= bs
->file
);
646 static inline bool data_file_is_raw(BlockDriverState
*bs
)
648 BDRVQcow2State
*s
= bs
->opaque
;
649 return !!(s
->autoclear_features
& QCOW2_AUTOCLEAR_DATA_FILE_RAW
);
652 static inline int64_t start_of_cluster(BDRVQcow2State
*s
, int64_t offset
)
654 return offset
& ~(s
->cluster_size
- 1);
657 static inline int64_t offset_into_cluster(BDRVQcow2State
*s
, int64_t offset
)
659 return offset
& (s
->cluster_size
- 1);
662 static inline int64_t offset_into_subcluster(BDRVQcow2State
*s
, int64_t offset
)
664 return offset
& (s
->subcluster_size
- 1);
667 static inline uint64_t size_to_clusters(BDRVQcow2State
*s
, uint64_t size
)
669 return (size
+ (s
->cluster_size
- 1)) >> s
->cluster_bits
;
672 static inline uint64_t size_to_subclusters(BDRVQcow2State
*s
, uint64_t size
)
674 return (size
+ (s
->subcluster_size
- 1)) >> s
->subcluster_bits
;
677 static inline int64_t size_to_l1(BDRVQcow2State
*s
, int64_t size
)
679 int shift
= s
->cluster_bits
+ s
->l2_bits
;
680 return (size
+ (1ULL << shift
) - 1) >> shift
;
683 static inline int offset_to_l1_index(BDRVQcow2State
*s
, uint64_t offset
)
685 return offset
>> (s
->l2_bits
+ s
->cluster_bits
);
688 static inline int offset_to_l2_index(BDRVQcow2State
*s
, int64_t offset
)
690 return (offset
>> s
->cluster_bits
) & (s
->l2_size
- 1);
693 static inline int offset_to_l2_slice_index(BDRVQcow2State
*s
, int64_t offset
)
695 return (offset
>> s
->cluster_bits
) & (s
->l2_slice_size
- 1);
698 static inline int offset_to_sc_index(BDRVQcow2State
*s
, int64_t offset
)
700 return (offset
>> s
->subcluster_bits
) & (s
->subclusters_per_cluster
- 1);
703 static inline int64_t qcow2_vm_state_offset(BDRVQcow2State
*s
)
705 return (int64_t)s
->l1_vm_state_index
<< (s
->cluster_bits
+ s
->l2_bits
);
708 static inline QCow2ClusterType
qcow2_get_cluster_type(BlockDriverState
*bs
,
711 BDRVQcow2State
*s
= bs
->opaque
;
713 if (l2_entry
& QCOW_OFLAG_COMPRESSED
) {
714 return QCOW2_CLUSTER_COMPRESSED
;
715 } else if ((l2_entry
& QCOW_OFLAG_ZERO
) && !has_subclusters(s
)) {
716 if (l2_entry
& L2E_OFFSET_MASK
) {
717 return QCOW2_CLUSTER_ZERO_ALLOC
;
719 return QCOW2_CLUSTER_ZERO_PLAIN
;
720 } else if (!(l2_entry
& L2E_OFFSET_MASK
)) {
721 /* Offset 0 generally means unallocated, but it is ambiguous with
722 * external data files because 0 is a valid offset there. However, all
723 * clusters in external data files always have refcount 1, so we can
724 * rely on QCOW_OFLAG_COPIED to disambiguate. */
725 if (has_data_file(bs
) && (l2_entry
& QCOW_OFLAG_COPIED
)) {
726 return QCOW2_CLUSTER_NORMAL
;
728 return QCOW2_CLUSTER_UNALLOCATED
;
731 return QCOW2_CLUSTER_NORMAL
;
736 * In an image without subsclusters @l2_bitmap is ignored and
737 * @sc_index must be 0.
738 * Return QCOW2_SUBCLUSTER_INVALID if an invalid l2 entry is detected
739 * (this checks the whole entry and bitmap, not only the bits related
740 * to subcluster @sc_index).
743 QCow2SubclusterType
qcow2_get_subcluster_type(BlockDriverState
*bs
,
748 BDRVQcow2State
*s
= bs
->opaque
;
749 QCow2ClusterType type
= qcow2_get_cluster_type(bs
, l2_entry
);
750 assert(sc_index
< s
->subclusters_per_cluster
);
752 if (has_subclusters(s
)) {
754 case QCOW2_CLUSTER_COMPRESSED
:
755 return QCOW2_SUBCLUSTER_COMPRESSED
;
756 case QCOW2_CLUSTER_NORMAL
:
757 if ((l2_bitmap
>> 32) & l2_bitmap
) {
758 return QCOW2_SUBCLUSTER_INVALID
;
759 } else if (l2_bitmap
& QCOW_OFLAG_SUB_ZERO(sc_index
)) {
760 return QCOW2_SUBCLUSTER_ZERO_ALLOC
;
761 } else if (l2_bitmap
& QCOW_OFLAG_SUB_ALLOC(sc_index
)) {
762 return QCOW2_SUBCLUSTER_NORMAL
;
764 return QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC
;
766 case QCOW2_CLUSTER_UNALLOCATED
:
767 if (l2_bitmap
& QCOW_L2_BITMAP_ALL_ALLOC
) {
768 return QCOW2_SUBCLUSTER_INVALID
;
769 } else if (l2_bitmap
& QCOW_OFLAG_SUB_ZERO(sc_index
)) {
770 return QCOW2_SUBCLUSTER_ZERO_PLAIN
;
772 return QCOW2_SUBCLUSTER_UNALLOCATED_PLAIN
;
775 g_assert_not_reached();
779 case QCOW2_CLUSTER_COMPRESSED
:
780 return QCOW2_SUBCLUSTER_COMPRESSED
;
781 case QCOW2_CLUSTER_ZERO_PLAIN
:
782 return QCOW2_SUBCLUSTER_ZERO_PLAIN
;
783 case QCOW2_CLUSTER_ZERO_ALLOC
:
784 return QCOW2_SUBCLUSTER_ZERO_ALLOC
;
785 case QCOW2_CLUSTER_NORMAL
:
786 return QCOW2_SUBCLUSTER_NORMAL
;
787 case QCOW2_CLUSTER_UNALLOCATED
:
788 return QCOW2_SUBCLUSTER_UNALLOCATED_PLAIN
;
790 g_assert_not_reached();
795 static inline bool qcow2_cluster_is_allocated(QCow2ClusterType type
)
797 return (type
== QCOW2_CLUSTER_COMPRESSED
|| type
== QCOW2_CLUSTER_NORMAL
||
798 type
== QCOW2_CLUSTER_ZERO_ALLOC
);
801 /* Check whether refcounts are eager or lazy */
802 static inline bool qcow2_need_accurate_refcounts(BDRVQcow2State
*s
)
804 return !(s
->incompatible_features
& QCOW2_INCOMPAT_DIRTY
);
807 static inline uint64_t l2meta_cow_start(QCowL2Meta
*m
)
809 return m
->offset
+ m
->cow_start
.offset
;
812 static inline uint64_t l2meta_cow_end(QCowL2Meta
*m
)
814 return m
->offset
+ m
->cow_end
.offset
+ m
->cow_end
.nb_bytes
;
817 static inline uint64_t refcount_diff(uint64_t r1
, uint64_t r2
)
819 return r1
> r2
? r1
- r2
: r2
- r1
;
823 uint32_t offset_to_reftable_index(BDRVQcow2State
*s
, uint64_t offset
)
825 return offset
>> (s
->refcount_block_bits
+ s
->cluster_bits
);
828 /* qcow2.c functions */
829 int64_t qcow2_refcount_metadata_size(int64_t clusters
, size_t cluster_size
,
830 int refcount_order
, bool generous_increase
,
831 uint64_t *refblock_count
);
833 int qcow2_mark_dirty(BlockDriverState
*bs
);
834 int qcow2_mark_corrupt(BlockDriverState
*bs
);
835 int qcow2_mark_consistent(BlockDriverState
*bs
);
836 int qcow2_update_header(BlockDriverState
*bs
);
838 void qcow2_signal_corruption(BlockDriverState
*bs
, bool fatal
, int64_t offset
,
839 int64_t size
, const char *message_format
, ...)
842 int qcow2_validate_table(BlockDriverState
*bs
, uint64_t offset
,
843 uint64_t entries
, size_t entry_len
,
844 int64_t max_size_bytes
, const char *table_name
,
847 /* qcow2-refcount.c functions */
848 int qcow2_refcount_init(BlockDriverState
*bs
);
849 void qcow2_refcount_close(BlockDriverState
*bs
);
851 int qcow2_get_refcount(BlockDriverState
*bs
, int64_t cluster_index
,
854 int qcow2_update_cluster_refcount(BlockDriverState
*bs
, int64_t cluster_index
,
855 uint64_t addend
, bool decrease
,
856 enum qcow2_discard_type type
);
858 int64_t qcow2_refcount_area(BlockDriverState
*bs
, uint64_t offset
,
859 uint64_t additional_clusters
, bool exact_size
,
860 int new_refblock_index
,
861 uint64_t new_refblock_offset
);
863 int64_t qcow2_alloc_clusters(BlockDriverState
*bs
, uint64_t size
);
864 int64_t qcow2_alloc_clusters_at(BlockDriverState
*bs
, uint64_t offset
,
865 int64_t nb_clusters
);
866 int64_t qcow2_alloc_bytes(BlockDriverState
*bs
, int size
);
867 void qcow2_free_clusters(BlockDriverState
*bs
,
868 int64_t offset
, int64_t size
,
869 enum qcow2_discard_type type
);
870 void qcow2_free_any_cluster(BlockDriverState
*bs
, uint64_t l2_entry
,
871 enum qcow2_discard_type type
);
873 int qcow2_update_snapshot_refcount(BlockDriverState
*bs
,
874 int64_t l1_table_offset
, int l1_size
, int addend
);
876 int coroutine_fn
qcow2_flush_caches(BlockDriverState
*bs
);
877 int coroutine_fn
qcow2_write_caches(BlockDriverState
*bs
);
878 int qcow2_check_refcounts(BlockDriverState
*bs
, BdrvCheckResult
*res
,
881 void qcow2_process_discards(BlockDriverState
*bs
, int ret
);
883 int qcow2_check_metadata_overlap(BlockDriverState
*bs
, int ign
, int64_t offset
,
885 int qcow2_pre_write_overlap_check(BlockDriverState
*bs
, int ign
, int64_t offset
,
886 int64_t size
, bool data_file
);
887 int qcow2_inc_refcounts_imrt(BlockDriverState
*bs
, BdrvCheckResult
*res
,
888 void **refcount_table
,
889 int64_t *refcount_table_size
,
890 int64_t offset
, int64_t size
);
892 int qcow2_change_refcount_order(BlockDriverState
*bs
, int refcount_order
,
893 BlockDriverAmendStatusCB
*status_cb
,
894 void *cb_opaque
, Error
**errp
);
895 int qcow2_shrink_reftable(BlockDriverState
*bs
);
896 int64_t qcow2_get_last_cluster(BlockDriverState
*bs
, int64_t size
);
897 int qcow2_detect_metadata_preallocation(BlockDriverState
*bs
);
899 /* qcow2-cluster.c functions */
900 int qcow2_grow_l1_table(BlockDriverState
*bs
, uint64_t min_size
,
902 int qcow2_shrink_l1_table(BlockDriverState
*bs
, uint64_t max_size
);
903 int qcow2_write_l1_entry(BlockDriverState
*bs
, int l1_index
);
904 int qcow2_encrypt_sectors(BDRVQcow2State
*s
, int64_t sector_num
,
905 uint8_t *buf
, int nb_sectors
, bool enc
, Error
**errp
);
907 int qcow2_get_host_offset(BlockDriverState
*bs
, uint64_t offset
,
908 unsigned int *bytes
, uint64_t *host_offset
,
909 QCow2SubclusterType
*subcluster_type
);
910 int qcow2_alloc_host_offset(BlockDriverState
*bs
, uint64_t offset
,
911 unsigned int *bytes
, uint64_t *host_offset
,
913 int qcow2_alloc_compressed_cluster_offset(BlockDriverState
*bs
,
916 uint64_t *host_offset
);
918 int qcow2_alloc_cluster_link_l2(BlockDriverState
*bs
, QCowL2Meta
*m
);
919 void qcow2_alloc_cluster_abort(BlockDriverState
*bs
, QCowL2Meta
*m
);
920 int qcow2_cluster_discard(BlockDriverState
*bs
, uint64_t offset
,
921 uint64_t bytes
, enum qcow2_discard_type type
,
923 int qcow2_subcluster_zeroize(BlockDriverState
*bs
, uint64_t offset
,
924 uint64_t bytes
, int flags
);
926 int qcow2_expand_zero_clusters(BlockDriverState
*bs
,
927 BlockDriverAmendStatusCB
*status_cb
,
930 /* qcow2-snapshot.c functions */
931 int qcow2_snapshot_create(BlockDriverState
*bs
, QEMUSnapshotInfo
*sn_info
);
932 int qcow2_snapshot_goto(BlockDriverState
*bs
, const char *snapshot_id
);
933 int qcow2_snapshot_delete(BlockDriverState
*bs
,
934 const char *snapshot_id
,
937 int qcow2_snapshot_list(BlockDriverState
*bs
, QEMUSnapshotInfo
**psn_tab
);
938 int qcow2_snapshot_load_tmp(BlockDriverState
*bs
,
939 const char *snapshot_id
,
943 void qcow2_free_snapshots(BlockDriverState
*bs
);
944 int qcow2_read_snapshots(BlockDriverState
*bs
, Error
**errp
);
945 int qcow2_write_snapshots(BlockDriverState
*bs
);
947 int coroutine_fn
qcow2_check_read_snapshot_table(BlockDriverState
*bs
,
948 BdrvCheckResult
*result
,
950 int coroutine_fn
qcow2_check_fix_snapshot_table(BlockDriverState
*bs
,
951 BdrvCheckResult
*result
,
954 /* qcow2-cache.c functions */
955 Qcow2Cache
*qcow2_cache_create(BlockDriverState
*bs
, int num_tables
,
956 unsigned table_size
);
957 int qcow2_cache_destroy(Qcow2Cache
*c
);
959 void qcow2_cache_entry_mark_dirty(Qcow2Cache
*c
, void *table
);
960 int qcow2_cache_flush(BlockDriverState
*bs
, Qcow2Cache
*c
);
961 int qcow2_cache_write(BlockDriverState
*bs
, Qcow2Cache
*c
);
962 int qcow2_cache_set_dependency(BlockDriverState
*bs
, Qcow2Cache
*c
,
963 Qcow2Cache
*dependency
);
964 void qcow2_cache_depends_on_flush(Qcow2Cache
*c
);
966 void qcow2_cache_clean_unused(Qcow2Cache
*c
);
967 int qcow2_cache_empty(BlockDriverState
*bs
, Qcow2Cache
*c
);
969 int qcow2_cache_get(BlockDriverState
*bs
, Qcow2Cache
*c
, uint64_t offset
,
971 int qcow2_cache_get_empty(BlockDriverState
*bs
, Qcow2Cache
*c
, uint64_t offset
,
973 void qcow2_cache_put(Qcow2Cache
*c
, void **table
);
974 void *qcow2_cache_is_table_offset(Qcow2Cache
*c
, uint64_t offset
);
975 void qcow2_cache_discard(Qcow2Cache
*c
, void *table
);
977 /* qcow2-bitmap.c functions */
978 int qcow2_check_bitmaps_refcounts(BlockDriverState
*bs
, BdrvCheckResult
*res
,
979 void **refcount_table
,
980 int64_t *refcount_table_size
);
981 bool qcow2_load_dirty_bitmaps(BlockDriverState
*bs
, Error
**errp
);
982 Qcow2BitmapInfoList
*qcow2_get_bitmap_info_list(BlockDriverState
*bs
,
984 int qcow2_reopen_bitmaps_rw(BlockDriverState
*bs
, Error
**errp
);
985 int qcow2_truncate_bitmaps_check(BlockDriverState
*bs
, Error
**errp
);
986 void qcow2_store_persistent_dirty_bitmaps(BlockDriverState
*bs
,
987 bool release_stored
, Error
**errp
);
988 int qcow2_reopen_bitmaps_ro(BlockDriverState
*bs
, Error
**errp
);
989 bool qcow2_co_can_store_new_dirty_bitmap(BlockDriverState
*bs
,
991 uint32_t granularity
,
993 int qcow2_co_remove_persistent_dirty_bitmap(BlockDriverState
*bs
,
996 bool qcow2_supports_persistent_dirty_bitmap(BlockDriverState
*bs
);
997 uint64_t qcow2_get_persistent_dirty_bitmap_size(BlockDriverState
*bs
,
998 uint32_t cluster_size
);
1000 ssize_t coroutine_fn
1001 qcow2_co_compress(BlockDriverState
*bs
, void *dest
, size_t dest_size
,
1002 const void *src
, size_t src_size
);
1003 ssize_t coroutine_fn
1004 qcow2_co_decompress(BlockDriverState
*bs
, void *dest
, size_t dest_size
,
1005 const void *src
, size_t src_size
);
1007 qcow2_co_encrypt(BlockDriverState
*bs
, uint64_t host_offset
,
1008 uint64_t guest_offset
, void *buf
, size_t len
);
1010 qcow2_co_decrypt(BlockDriverState
*bs
, uint64_t host_offset
,
1011 uint64_t guest_offset
, void *buf
, size_t len
);