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
29 #include "block/coroutine.h"
32 //#define DEBUG_ALLOC2
35 #define QCOW_MAGIC (('Q' << 24) | ('F' << 16) | ('I' << 8) | 0xfb)
37 #define QCOW_CRYPT_NONE 0
38 #define QCOW_CRYPT_AES 1
40 #define QCOW_MAX_CRYPT_CLUSTERS 32
41 #define QCOW_MAX_SNAPSHOTS 65536
43 /* 8 MB refcount table is enough for 2 PB images at 64k cluster size
44 * (128 GB for 512 byte clusters, 2 EB for 2 MB clusters) */
45 #define QCOW_MAX_REFTABLE_SIZE 0x800000
47 /* indicate that the refcount of the referenced cluster is exactly one. */
48 #define QCOW_OFLAG_COPIED (1ULL << 63)
49 /* indicate that the cluster is compressed (they never have the copied flag) */
50 #define QCOW_OFLAG_COMPRESSED (1ULL << 62)
51 /* The cluster reads as all zeros */
52 #define QCOW_OFLAG_ZERO (1ULL << 0)
54 #define REFCOUNT_SHIFT 1 /* refcount size is 2 bytes */
56 #define MIN_CLUSTER_BITS 9
57 #define MAX_CLUSTER_BITS 21
59 #define L2_CACHE_SIZE 16
61 /* Must be at least 4 to cover all cases of refcount table growth */
62 #define REFCOUNT_CACHE_SIZE 4
64 #define DEFAULT_CLUSTER_SIZE 65536
67 #define QCOW2_OPT_LAZY_REFCOUNTS "lazy-refcounts"
68 #define QCOW2_OPT_DISCARD_REQUEST "pass-discard-request"
69 #define QCOW2_OPT_DISCARD_SNAPSHOT "pass-discard-snapshot"
70 #define QCOW2_OPT_DISCARD_OTHER "pass-discard-other"
71 #define QCOW2_OPT_OVERLAP "overlap-check"
72 #define QCOW2_OPT_OVERLAP_MAIN_HEADER "overlap-check.main-header"
73 #define QCOW2_OPT_OVERLAP_ACTIVE_L1 "overlap-check.active-l1"
74 #define QCOW2_OPT_OVERLAP_ACTIVE_L2 "overlap-check.active-l2"
75 #define QCOW2_OPT_OVERLAP_REFCOUNT_TABLE "overlap-check.refcount-table"
76 #define QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK "overlap-check.refcount-block"
77 #define QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE "overlap-check.snapshot-table"
78 #define QCOW2_OPT_OVERLAP_INACTIVE_L1 "overlap-check.inactive-l1"
79 #define QCOW2_OPT_OVERLAP_INACTIVE_L2 "overlap-check.inactive-l2"
81 typedef struct QCowHeader
{
84 uint64_t backing_file_offset
;
85 uint32_t backing_file_size
;
86 uint32_t cluster_bits
;
87 uint64_t size
; /* in bytes */
88 uint32_t crypt_method
;
89 uint32_t l1_size
; /* XXX: save number of clusters instead ? */
90 uint64_t l1_table_offset
;
91 uint64_t refcount_table_offset
;
92 uint32_t refcount_table_clusters
;
93 uint32_t nb_snapshots
;
94 uint64_t snapshots_offset
;
96 /* The following fields are only valid for version >= 3 */
97 uint64_t incompatible_features
;
98 uint64_t compatible_features
;
99 uint64_t autoclear_features
;
101 uint32_t refcount_order
;
102 uint32_t header_length
;
103 } QEMU_PACKED QCowHeader
;
105 typedef struct QEMU_PACKED QCowSnapshotHeader
{
106 /* header is 8 byte aligned */
107 uint64_t l1_table_offset
;
110 uint16_t id_str_size
;
116 uint64_t vm_clock_nsec
;
118 uint32_t vm_state_size
;
119 uint32_t extra_data_size
; /* for extension */
120 /* extra data follows */
123 } QCowSnapshotHeader
;
125 typedef struct QEMU_PACKED QCowSnapshotExtraData
{
126 uint64_t vm_state_size_large
;
128 } QCowSnapshotExtraData
;
131 typedef struct QCowSnapshot
{
132 uint64_t l1_table_offset
;
137 uint64_t vm_state_size
;
140 uint64_t vm_clock_nsec
;
144 typedef struct Qcow2Cache Qcow2Cache
;
146 typedef struct Qcow2UnknownHeaderExtension
{
149 QLIST_ENTRY(Qcow2UnknownHeaderExtension
) next
;
151 } Qcow2UnknownHeaderExtension
;
154 QCOW2_FEAT_TYPE_INCOMPATIBLE
= 0,
155 QCOW2_FEAT_TYPE_COMPATIBLE
= 1,
156 QCOW2_FEAT_TYPE_AUTOCLEAR
= 2,
159 /* Incompatible feature bits */
161 QCOW2_INCOMPAT_DIRTY_BITNR
= 0,
162 QCOW2_INCOMPAT_CORRUPT_BITNR
= 1,
163 QCOW2_INCOMPAT_DIRTY
= 1 << QCOW2_INCOMPAT_DIRTY_BITNR
,
164 QCOW2_INCOMPAT_CORRUPT
= 1 << QCOW2_INCOMPAT_CORRUPT_BITNR
,
166 QCOW2_INCOMPAT_MASK
= QCOW2_INCOMPAT_DIRTY
167 | QCOW2_INCOMPAT_CORRUPT
,
170 /* Compatible feature bits */
172 QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR
= 0,
173 QCOW2_COMPAT_LAZY_REFCOUNTS
= 1 << QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR
,
175 QCOW2_COMPAT_FEAT_MASK
= QCOW2_COMPAT_LAZY_REFCOUNTS
,
178 enum qcow2_discard_type
{
179 QCOW2_DISCARD_NEVER
= 0,
180 QCOW2_DISCARD_ALWAYS
,
181 QCOW2_DISCARD_REQUEST
,
182 QCOW2_DISCARD_SNAPSHOT
,
187 typedef struct Qcow2Feature
{
191 } QEMU_PACKED Qcow2Feature
;
193 typedef struct Qcow2DiscardRegion
{
194 BlockDriverState
*bs
;
197 QTAILQ_ENTRY(Qcow2DiscardRegion
) next
;
198 } Qcow2DiscardRegion
;
200 typedef struct BDRVQcowState
{
207 int l1_vm_state_index
;
210 uint64_t cluster_offset_mask
;
211 uint64_t l1_table_offset
;
214 Qcow2Cache
* l2_table_cache
;
215 Qcow2Cache
* refcount_block_cache
;
217 uint8_t *cluster_cache
;
218 uint8_t *cluster_data
;
219 uint64_t cluster_cache_offset
;
220 QLIST_HEAD(QCowClusterAlloc
, QCowL2Meta
) cluster_allocs
;
222 uint64_t *refcount_table
;
223 uint64_t refcount_table_offset
;
224 uint32_t refcount_table_size
;
225 uint64_t free_cluster_index
;
226 uint64_t free_byte_offset
;
230 uint32_t crypt_method
; /* current crypt method, 0 if no key yet */
231 uint32_t crypt_method_header
;
232 AES_KEY aes_encrypt_key
;
233 AES_KEY aes_decrypt_key
;
234 uint64_t snapshots_offset
;
236 unsigned int nb_snapshots
;
237 QCowSnapshot
*snapshots
;
241 bool use_lazy_refcounts
;
244 bool discard_passthrough
[QCOW2_DISCARD_MAX
];
246 int overlap_check
; /* bitmask of Qcow2MetadataOverlap values */
248 uint64_t incompatible_features
;
249 uint64_t compatible_features
;
250 uint64_t autoclear_features
;
252 size_t unknown_header_fields_size
;
253 void* unknown_header_fields
;
254 QLIST_HEAD(, Qcow2UnknownHeaderExtension
) unknown_header_ext
;
255 QTAILQ_HEAD (, Qcow2DiscardRegion
) discards
;
259 /* XXX: use std qcow open function ? */
260 typedef struct QCowCreateState
{
263 uint16_t *refcount_block
;
264 uint64_t *refcount_table
;
265 int64_t l1_table_offset
;
266 int64_t refcount_table_offset
;
267 int64_t refcount_block_offset
;
272 typedef struct Qcow2COWRegion
{
274 * Offset of the COW region in bytes from the start of the first cluster
275 * touched by the request.
279 /** Number of sectors to copy */
284 * Describes an in-flight (part of a) write request that writes to clusters
285 * that are not referenced in their L2 table yet.
287 typedef struct QCowL2Meta
289 /** Guest offset of the first newly allocated cluster */
292 /** Host offset of the first newly allocated cluster */
293 uint64_t alloc_offset
;
296 * Number of sectors from the start of the first allocated cluster to
297 * the end of the (possibly shortened) request
301 /** Number of newly allocated clusters */
305 * Requests that overlap with this allocation and wait to be restarted
306 * when the allocating request has completed.
308 CoQueue dependent_requests
;
311 * The COW Region between the start of the first allocated cluster and the
312 * area the guest actually writes to.
314 Qcow2COWRegion cow_start
;
317 * The COW Region between the area the guest actually writes to and the
318 * end of the last allocated cluster.
320 Qcow2COWRegion cow_end
;
322 /** Pointer to next L2Meta of the same write request */
323 struct QCowL2Meta
*next
;
325 QLIST_ENTRY(QCowL2Meta
) next_in_flight
;
329 QCOW2_CLUSTER_UNALLOCATED
,
330 QCOW2_CLUSTER_NORMAL
,
331 QCOW2_CLUSTER_COMPRESSED
,
335 typedef enum QCow2MetadataOverlap
{
336 QCOW2_OL_MAIN_HEADER_BITNR
= 0,
337 QCOW2_OL_ACTIVE_L1_BITNR
= 1,
338 QCOW2_OL_ACTIVE_L2_BITNR
= 2,
339 QCOW2_OL_REFCOUNT_TABLE_BITNR
= 3,
340 QCOW2_OL_REFCOUNT_BLOCK_BITNR
= 4,
341 QCOW2_OL_SNAPSHOT_TABLE_BITNR
= 5,
342 QCOW2_OL_INACTIVE_L1_BITNR
= 6,
343 QCOW2_OL_INACTIVE_L2_BITNR
= 7,
345 QCOW2_OL_MAX_BITNR
= 8,
348 QCOW2_OL_MAIN_HEADER
= (1 << QCOW2_OL_MAIN_HEADER_BITNR
),
349 QCOW2_OL_ACTIVE_L1
= (1 << QCOW2_OL_ACTIVE_L1_BITNR
),
350 QCOW2_OL_ACTIVE_L2
= (1 << QCOW2_OL_ACTIVE_L2_BITNR
),
351 QCOW2_OL_REFCOUNT_TABLE
= (1 << QCOW2_OL_REFCOUNT_TABLE_BITNR
),
352 QCOW2_OL_REFCOUNT_BLOCK
= (1 << QCOW2_OL_REFCOUNT_BLOCK_BITNR
),
353 QCOW2_OL_SNAPSHOT_TABLE
= (1 << QCOW2_OL_SNAPSHOT_TABLE_BITNR
),
354 QCOW2_OL_INACTIVE_L1
= (1 << QCOW2_OL_INACTIVE_L1_BITNR
),
355 /* NOTE: Checking overlaps with inactive L2 tables will result in bdrv
357 QCOW2_OL_INACTIVE_L2
= (1 << QCOW2_OL_INACTIVE_L2_BITNR
),
358 } QCow2MetadataOverlap
;
360 /* Perform all overlap checks which can be done in constant time */
361 #define QCOW2_OL_CONSTANT \
362 (QCOW2_OL_MAIN_HEADER | QCOW2_OL_ACTIVE_L1 | QCOW2_OL_REFCOUNT_TABLE | \
363 QCOW2_OL_SNAPSHOT_TABLE)
365 /* Perform all overlap checks which don't require disk access */
366 #define QCOW2_OL_CACHED \
367 (QCOW2_OL_CONSTANT | QCOW2_OL_ACTIVE_L2 | QCOW2_OL_REFCOUNT_BLOCK | \
368 QCOW2_OL_INACTIVE_L1)
370 /* Perform all overlap checks */
371 #define QCOW2_OL_ALL \
372 (QCOW2_OL_CACHED | QCOW2_OL_INACTIVE_L2)
374 #define L1E_OFFSET_MASK 0x00fffffffffffe00ULL
375 #define L2E_OFFSET_MASK 0x00fffffffffffe00ULL
376 #define L2E_COMPRESSED_OFFSET_SIZE_MASK 0x3fffffffffffffffULL
378 #define REFT_OFFSET_MASK 0xfffffffffffffe00ULL
380 static inline int64_t start_of_cluster(BDRVQcowState
*s
, int64_t offset
)
382 return offset
& ~(s
->cluster_size
- 1);
385 static inline int64_t offset_into_cluster(BDRVQcowState
*s
, int64_t offset
)
387 return offset
& (s
->cluster_size
- 1);
390 static inline int size_to_clusters(BDRVQcowState
*s
, int64_t size
)
392 return (size
+ (s
->cluster_size
- 1)) >> s
->cluster_bits
;
395 static inline int64_t size_to_l1(BDRVQcowState
*s
, int64_t size
)
397 int shift
= s
->cluster_bits
+ s
->l2_bits
;
398 return (size
+ (1ULL << shift
) - 1) >> shift
;
401 static inline int offset_to_l2_index(BDRVQcowState
*s
, int64_t offset
)
403 return (offset
>> s
->cluster_bits
) & (s
->l2_size
- 1);
406 static inline int64_t align_offset(int64_t offset
, int n
)
408 offset
= (offset
+ n
- 1) & ~(n
- 1);
412 static inline int64_t qcow2_vm_state_offset(BDRVQcowState
*s
)
414 return (int64_t)s
->l1_vm_state_index
<< (s
->cluster_bits
+ s
->l2_bits
);
417 static inline uint64_t qcow2_max_refcount_clusters(BDRVQcowState
*s
)
419 return QCOW_MAX_REFTABLE_SIZE
>> s
->cluster_bits
;
422 static inline int qcow2_get_cluster_type(uint64_t l2_entry
)
424 if (l2_entry
& QCOW_OFLAG_COMPRESSED
) {
425 return QCOW2_CLUSTER_COMPRESSED
;
426 } else if (l2_entry
& QCOW_OFLAG_ZERO
) {
427 return QCOW2_CLUSTER_ZERO
;
428 } else if (!(l2_entry
& L2E_OFFSET_MASK
)) {
429 return QCOW2_CLUSTER_UNALLOCATED
;
431 return QCOW2_CLUSTER_NORMAL
;
435 /* Check whether refcounts are eager or lazy */
436 static inline bool qcow2_need_accurate_refcounts(BDRVQcowState
*s
)
438 return !(s
->incompatible_features
& QCOW2_INCOMPAT_DIRTY
);
441 static inline uint64_t l2meta_cow_start(QCowL2Meta
*m
)
443 return m
->offset
+ m
->cow_start
.offset
;
446 static inline uint64_t l2meta_cow_end(QCowL2Meta
*m
)
448 return m
->offset
+ m
->cow_end
.offset
449 + (m
->cow_end
.nb_sectors
<< BDRV_SECTOR_BITS
);
452 // FIXME Need qcow2_ prefix to global functions
454 /* qcow2.c functions */
455 int qcow2_backing_read1(BlockDriverState
*bs
, QEMUIOVector
*qiov
,
456 int64_t sector_num
, int nb_sectors
);
458 int qcow2_mark_dirty(BlockDriverState
*bs
);
459 int qcow2_mark_corrupt(BlockDriverState
*bs
);
460 int qcow2_mark_consistent(BlockDriverState
*bs
);
461 int qcow2_update_header(BlockDriverState
*bs
);
463 /* qcow2-refcount.c functions */
464 int qcow2_refcount_init(BlockDriverState
*bs
);
465 void qcow2_refcount_close(BlockDriverState
*bs
);
467 int qcow2_update_cluster_refcount(BlockDriverState
*bs
, int64_t cluster_index
,
468 int addend
, enum qcow2_discard_type type
);
470 int64_t qcow2_alloc_clusters(BlockDriverState
*bs
, uint64_t size
);
471 int qcow2_alloc_clusters_at(BlockDriverState
*bs
, uint64_t offset
,
473 int64_t qcow2_alloc_bytes(BlockDriverState
*bs
, int size
);
474 void qcow2_free_clusters(BlockDriverState
*bs
,
475 int64_t offset
, int64_t size
,
476 enum qcow2_discard_type type
);
477 void qcow2_free_any_clusters(BlockDriverState
*bs
, uint64_t l2_entry
,
478 int nb_clusters
, enum qcow2_discard_type type
);
480 int qcow2_update_snapshot_refcount(BlockDriverState
*bs
,
481 int64_t l1_table_offset
, int l1_size
, int addend
);
483 int qcow2_check_refcounts(BlockDriverState
*bs
, BdrvCheckResult
*res
,
486 void qcow2_process_discards(BlockDriverState
*bs
, int ret
);
488 int qcow2_check_metadata_overlap(BlockDriverState
*bs
, int ign
, int64_t offset
,
490 int qcow2_pre_write_overlap_check(BlockDriverState
*bs
, int ign
, int64_t offset
,
493 /* qcow2-cluster.c functions */
494 int qcow2_grow_l1_table(BlockDriverState
*bs
, uint64_t min_size
,
496 int qcow2_write_l1_entry(BlockDriverState
*bs
, int l1_index
);
497 void qcow2_l2_cache_reset(BlockDriverState
*bs
);
498 int qcow2_decompress_cluster(BlockDriverState
*bs
, uint64_t cluster_offset
);
499 void qcow2_encrypt_sectors(BDRVQcowState
*s
, int64_t sector_num
,
500 uint8_t *out_buf
, const uint8_t *in_buf
,
501 int nb_sectors
, int enc
,
504 int qcow2_get_cluster_offset(BlockDriverState
*bs
, uint64_t offset
,
505 int *num
, uint64_t *cluster_offset
);
506 int qcow2_alloc_cluster_offset(BlockDriverState
*bs
, uint64_t offset
,
507 int *num
, uint64_t *host_offset
, QCowL2Meta
**m
);
508 uint64_t qcow2_alloc_compressed_cluster_offset(BlockDriverState
*bs
,
510 int compressed_size
);
512 int qcow2_alloc_cluster_link_l2(BlockDriverState
*bs
, QCowL2Meta
*m
);
513 int qcow2_discard_clusters(BlockDriverState
*bs
, uint64_t offset
,
514 int nb_sectors
, enum qcow2_discard_type type
);
515 int qcow2_zero_clusters(BlockDriverState
*bs
, uint64_t offset
, int nb_sectors
);
517 int qcow2_expand_zero_clusters(BlockDriverState
*bs
);
519 /* qcow2-snapshot.c functions */
520 int qcow2_snapshot_create(BlockDriverState
*bs
, QEMUSnapshotInfo
*sn_info
);
521 int qcow2_snapshot_goto(BlockDriverState
*bs
, const char *snapshot_id
);
522 int qcow2_snapshot_delete(BlockDriverState
*bs
,
523 const char *snapshot_id
,
526 int qcow2_snapshot_list(BlockDriverState
*bs
, QEMUSnapshotInfo
**psn_tab
);
527 int qcow2_snapshot_load_tmp(BlockDriverState
*bs
,
528 const char *snapshot_id
,
532 void qcow2_free_snapshots(BlockDriverState
*bs
);
533 int qcow2_read_snapshots(BlockDriverState
*bs
);
535 /* qcow2-cache.c functions */
536 Qcow2Cache
*qcow2_cache_create(BlockDriverState
*bs
, int num_tables
);
537 int qcow2_cache_destroy(BlockDriverState
* bs
, Qcow2Cache
*c
);
539 void qcow2_cache_entry_mark_dirty(Qcow2Cache
*c
, void *table
);
540 int qcow2_cache_flush(BlockDriverState
*bs
, Qcow2Cache
*c
);
541 int qcow2_cache_set_dependency(BlockDriverState
*bs
, Qcow2Cache
*c
,
542 Qcow2Cache
*dependency
);
543 void qcow2_cache_depends_on_flush(Qcow2Cache
*c
);
545 int qcow2_cache_empty(BlockDriverState
*bs
, Qcow2Cache
*c
);
547 int qcow2_cache_get(BlockDriverState
*bs
, Qcow2Cache
*c
, uint64_t offset
,
549 int qcow2_cache_get_empty(BlockDriverState
*bs
, Qcow2Cache
*c
, uint64_t offset
,
551 int qcow2_cache_put(BlockDriverState
*bs
, Qcow2Cache
*c
, void **table
);