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 /* 32 MB L1 table is enough for 2 PB images at 64k cluster size
48 * (128 GB for 512 byte clusters, 2 EB for 2 MB clusters) */
49 #define QCOW_MAX_L1_SIZE 0x2000000
51 /* Allow for an average of 1k per snapshot table entry, should be plenty of
52 * space for snapshot names and IDs */
53 #define QCOW_MAX_SNAPSHOTS_SIZE (1024 * QCOW_MAX_SNAPSHOTS)
55 /* indicate that the refcount of the referenced cluster is exactly one. */
56 #define QCOW_OFLAG_COPIED (1ULL << 63)
57 /* indicate that the cluster is compressed (they never have the copied flag) */
58 #define QCOW_OFLAG_COMPRESSED (1ULL << 62)
59 /* The cluster reads as all zeros */
60 #define QCOW_OFLAG_ZERO (1ULL << 0)
62 #define REFCOUNT_SHIFT 1 /* refcount size is 2 bytes */
64 #define MIN_CLUSTER_BITS 9
65 #define MAX_CLUSTER_BITS 21
67 #define L2_CACHE_SIZE 16
69 /* Must be at least 4 to cover all cases of refcount table growth */
70 #define REFCOUNT_CACHE_SIZE 4
72 #define DEFAULT_CLUSTER_SIZE 65536
75 #define QCOW2_OPT_LAZY_REFCOUNTS "lazy-refcounts"
76 #define QCOW2_OPT_DISCARD_REQUEST "pass-discard-request"
77 #define QCOW2_OPT_DISCARD_SNAPSHOT "pass-discard-snapshot"
78 #define QCOW2_OPT_DISCARD_OTHER "pass-discard-other"
79 #define QCOW2_OPT_OVERLAP "overlap-check"
80 #define QCOW2_OPT_OVERLAP_MAIN_HEADER "overlap-check.main-header"
81 #define QCOW2_OPT_OVERLAP_ACTIVE_L1 "overlap-check.active-l1"
82 #define QCOW2_OPT_OVERLAP_ACTIVE_L2 "overlap-check.active-l2"
83 #define QCOW2_OPT_OVERLAP_REFCOUNT_TABLE "overlap-check.refcount-table"
84 #define QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK "overlap-check.refcount-block"
85 #define QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE "overlap-check.snapshot-table"
86 #define QCOW2_OPT_OVERLAP_INACTIVE_L1 "overlap-check.inactive-l1"
87 #define QCOW2_OPT_OVERLAP_INACTIVE_L2 "overlap-check.inactive-l2"
89 typedef struct QCowHeader
{
92 uint64_t backing_file_offset
;
93 uint32_t backing_file_size
;
94 uint32_t cluster_bits
;
95 uint64_t size
; /* in bytes */
96 uint32_t crypt_method
;
97 uint32_t l1_size
; /* XXX: save number of clusters instead ? */
98 uint64_t l1_table_offset
;
99 uint64_t refcount_table_offset
;
100 uint32_t refcount_table_clusters
;
101 uint32_t nb_snapshots
;
102 uint64_t snapshots_offset
;
104 /* The following fields are only valid for version >= 3 */
105 uint64_t incompatible_features
;
106 uint64_t compatible_features
;
107 uint64_t autoclear_features
;
109 uint32_t refcount_order
;
110 uint32_t header_length
;
111 } QEMU_PACKED QCowHeader
;
113 typedef struct QEMU_PACKED QCowSnapshotHeader
{
114 /* header is 8 byte aligned */
115 uint64_t l1_table_offset
;
118 uint16_t id_str_size
;
124 uint64_t vm_clock_nsec
;
126 uint32_t vm_state_size
;
127 uint32_t extra_data_size
; /* for extension */
128 /* extra data follows */
131 } QCowSnapshotHeader
;
133 typedef struct QEMU_PACKED QCowSnapshotExtraData
{
134 uint64_t vm_state_size_large
;
136 } QCowSnapshotExtraData
;
139 typedef struct QCowSnapshot
{
140 uint64_t l1_table_offset
;
145 uint64_t vm_state_size
;
148 uint64_t vm_clock_nsec
;
152 typedef struct Qcow2Cache Qcow2Cache
;
154 typedef struct Qcow2UnknownHeaderExtension
{
157 QLIST_ENTRY(Qcow2UnknownHeaderExtension
) next
;
159 } Qcow2UnknownHeaderExtension
;
162 QCOW2_FEAT_TYPE_INCOMPATIBLE
= 0,
163 QCOW2_FEAT_TYPE_COMPATIBLE
= 1,
164 QCOW2_FEAT_TYPE_AUTOCLEAR
= 2,
167 /* Incompatible feature bits */
169 QCOW2_INCOMPAT_DIRTY_BITNR
= 0,
170 QCOW2_INCOMPAT_CORRUPT_BITNR
= 1,
171 QCOW2_INCOMPAT_DIRTY
= 1 << QCOW2_INCOMPAT_DIRTY_BITNR
,
172 QCOW2_INCOMPAT_CORRUPT
= 1 << QCOW2_INCOMPAT_CORRUPT_BITNR
,
174 QCOW2_INCOMPAT_MASK
= QCOW2_INCOMPAT_DIRTY
175 | QCOW2_INCOMPAT_CORRUPT
,
178 /* Compatible feature bits */
180 QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR
= 0,
181 QCOW2_COMPAT_LAZY_REFCOUNTS
= 1 << QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR
,
183 QCOW2_COMPAT_FEAT_MASK
= QCOW2_COMPAT_LAZY_REFCOUNTS
,
186 enum qcow2_discard_type
{
187 QCOW2_DISCARD_NEVER
= 0,
188 QCOW2_DISCARD_ALWAYS
,
189 QCOW2_DISCARD_REQUEST
,
190 QCOW2_DISCARD_SNAPSHOT
,
195 typedef struct Qcow2Feature
{
199 } QEMU_PACKED Qcow2Feature
;
201 typedef struct Qcow2DiscardRegion
{
202 BlockDriverState
*bs
;
205 QTAILQ_ENTRY(Qcow2DiscardRegion
) next
;
206 } Qcow2DiscardRegion
;
208 typedef struct BDRVQcowState
{
215 int l1_vm_state_index
;
218 uint64_t cluster_offset_mask
;
219 uint64_t l1_table_offset
;
222 Qcow2Cache
* l2_table_cache
;
223 Qcow2Cache
* refcount_block_cache
;
225 uint8_t *cluster_cache
;
226 uint8_t *cluster_data
;
227 uint64_t cluster_cache_offset
;
228 QLIST_HEAD(QCowClusterAlloc
, QCowL2Meta
) cluster_allocs
;
230 uint64_t *refcount_table
;
231 uint64_t refcount_table_offset
;
232 uint32_t refcount_table_size
;
233 uint64_t free_cluster_index
;
234 uint64_t free_byte_offset
;
238 uint32_t crypt_method
; /* current crypt method, 0 if no key yet */
239 uint32_t crypt_method_header
;
240 AES_KEY aes_encrypt_key
;
241 AES_KEY aes_decrypt_key
;
242 uint64_t snapshots_offset
;
244 unsigned int nb_snapshots
;
245 QCowSnapshot
*snapshots
;
249 bool use_lazy_refcounts
;
252 bool discard_passthrough
[QCOW2_DISCARD_MAX
];
254 int overlap_check
; /* bitmask of Qcow2MetadataOverlap values */
256 uint64_t incompatible_features
;
257 uint64_t compatible_features
;
258 uint64_t autoclear_features
;
260 size_t unknown_header_fields_size
;
261 void* unknown_header_fields
;
262 QLIST_HEAD(, Qcow2UnknownHeaderExtension
) unknown_header_ext
;
263 QTAILQ_HEAD (, Qcow2DiscardRegion
) discards
;
267 /* XXX: use std qcow open function ? */
268 typedef struct QCowCreateState
{
271 uint16_t *refcount_block
;
272 uint64_t *refcount_table
;
273 int64_t l1_table_offset
;
274 int64_t refcount_table_offset
;
275 int64_t refcount_block_offset
;
280 typedef struct Qcow2COWRegion
{
282 * Offset of the COW region in bytes from the start of the first cluster
283 * touched by the request.
287 /** Number of sectors to copy */
292 * Describes an in-flight (part of a) write request that writes to clusters
293 * that are not referenced in their L2 table yet.
295 typedef struct QCowL2Meta
297 /** Guest offset of the first newly allocated cluster */
300 /** Host offset of the first newly allocated cluster */
301 uint64_t alloc_offset
;
304 * Number of sectors from the start of the first allocated cluster to
305 * the end of the (possibly shortened) request
309 /** Number of newly allocated clusters */
313 * Requests that overlap with this allocation and wait to be restarted
314 * when the allocating request has completed.
316 CoQueue dependent_requests
;
319 * The COW Region between the start of the first allocated cluster and the
320 * area the guest actually writes to.
322 Qcow2COWRegion cow_start
;
325 * The COW Region between the area the guest actually writes to and the
326 * end of the last allocated cluster.
328 Qcow2COWRegion cow_end
;
330 /** Pointer to next L2Meta of the same write request */
331 struct QCowL2Meta
*next
;
333 QLIST_ENTRY(QCowL2Meta
) next_in_flight
;
337 QCOW2_CLUSTER_UNALLOCATED
,
338 QCOW2_CLUSTER_NORMAL
,
339 QCOW2_CLUSTER_COMPRESSED
,
343 typedef enum QCow2MetadataOverlap
{
344 QCOW2_OL_MAIN_HEADER_BITNR
= 0,
345 QCOW2_OL_ACTIVE_L1_BITNR
= 1,
346 QCOW2_OL_ACTIVE_L2_BITNR
= 2,
347 QCOW2_OL_REFCOUNT_TABLE_BITNR
= 3,
348 QCOW2_OL_REFCOUNT_BLOCK_BITNR
= 4,
349 QCOW2_OL_SNAPSHOT_TABLE_BITNR
= 5,
350 QCOW2_OL_INACTIVE_L1_BITNR
= 6,
351 QCOW2_OL_INACTIVE_L2_BITNR
= 7,
353 QCOW2_OL_MAX_BITNR
= 8,
356 QCOW2_OL_MAIN_HEADER
= (1 << QCOW2_OL_MAIN_HEADER_BITNR
),
357 QCOW2_OL_ACTIVE_L1
= (1 << QCOW2_OL_ACTIVE_L1_BITNR
),
358 QCOW2_OL_ACTIVE_L2
= (1 << QCOW2_OL_ACTIVE_L2_BITNR
),
359 QCOW2_OL_REFCOUNT_TABLE
= (1 << QCOW2_OL_REFCOUNT_TABLE_BITNR
),
360 QCOW2_OL_REFCOUNT_BLOCK
= (1 << QCOW2_OL_REFCOUNT_BLOCK_BITNR
),
361 QCOW2_OL_SNAPSHOT_TABLE
= (1 << QCOW2_OL_SNAPSHOT_TABLE_BITNR
),
362 QCOW2_OL_INACTIVE_L1
= (1 << QCOW2_OL_INACTIVE_L1_BITNR
),
363 /* NOTE: Checking overlaps with inactive L2 tables will result in bdrv
365 QCOW2_OL_INACTIVE_L2
= (1 << QCOW2_OL_INACTIVE_L2_BITNR
),
366 } QCow2MetadataOverlap
;
368 /* Perform all overlap checks which can be done in constant time */
369 #define QCOW2_OL_CONSTANT \
370 (QCOW2_OL_MAIN_HEADER | QCOW2_OL_ACTIVE_L1 | QCOW2_OL_REFCOUNT_TABLE | \
371 QCOW2_OL_SNAPSHOT_TABLE)
373 /* Perform all overlap checks which don't require disk access */
374 #define QCOW2_OL_CACHED \
375 (QCOW2_OL_CONSTANT | QCOW2_OL_ACTIVE_L2 | QCOW2_OL_REFCOUNT_BLOCK | \
376 QCOW2_OL_INACTIVE_L1)
378 /* Perform all overlap checks */
379 #define QCOW2_OL_ALL \
380 (QCOW2_OL_CACHED | QCOW2_OL_INACTIVE_L2)
382 #define L1E_OFFSET_MASK 0x00fffffffffffe00ULL
383 #define L2E_OFFSET_MASK 0x00fffffffffffe00ULL
384 #define L2E_COMPRESSED_OFFSET_SIZE_MASK 0x3fffffffffffffffULL
386 #define REFT_OFFSET_MASK 0xfffffffffffffe00ULL
388 static inline int64_t start_of_cluster(BDRVQcowState
*s
, int64_t offset
)
390 return offset
& ~(s
->cluster_size
- 1);
393 static inline int64_t offset_into_cluster(BDRVQcowState
*s
, int64_t offset
)
395 return offset
& (s
->cluster_size
- 1);
398 static inline int size_to_clusters(BDRVQcowState
*s
, int64_t size
)
400 return (size
+ (s
->cluster_size
- 1)) >> s
->cluster_bits
;
403 static inline int64_t size_to_l1(BDRVQcowState
*s
, int64_t size
)
405 int shift
= s
->cluster_bits
+ s
->l2_bits
;
406 return (size
+ (1ULL << shift
) - 1) >> shift
;
409 static inline int offset_to_l2_index(BDRVQcowState
*s
, int64_t offset
)
411 return (offset
>> s
->cluster_bits
) & (s
->l2_size
- 1);
414 static inline int64_t align_offset(int64_t offset
, int n
)
416 offset
= (offset
+ n
- 1) & ~(n
- 1);
420 static inline int64_t qcow2_vm_state_offset(BDRVQcowState
*s
)
422 return (int64_t)s
->l1_vm_state_index
<< (s
->cluster_bits
+ s
->l2_bits
);
425 static inline uint64_t qcow2_max_refcount_clusters(BDRVQcowState
*s
)
427 return QCOW_MAX_REFTABLE_SIZE
>> s
->cluster_bits
;
430 static inline int qcow2_get_cluster_type(uint64_t l2_entry
)
432 if (l2_entry
& QCOW_OFLAG_COMPRESSED
) {
433 return QCOW2_CLUSTER_COMPRESSED
;
434 } else if (l2_entry
& QCOW_OFLAG_ZERO
) {
435 return QCOW2_CLUSTER_ZERO
;
436 } else if (!(l2_entry
& L2E_OFFSET_MASK
)) {
437 return QCOW2_CLUSTER_UNALLOCATED
;
439 return QCOW2_CLUSTER_NORMAL
;
443 /* Check whether refcounts are eager or lazy */
444 static inline bool qcow2_need_accurate_refcounts(BDRVQcowState
*s
)
446 return !(s
->incompatible_features
& QCOW2_INCOMPAT_DIRTY
);
449 static inline uint64_t l2meta_cow_start(QCowL2Meta
*m
)
451 return m
->offset
+ m
->cow_start
.offset
;
454 static inline uint64_t l2meta_cow_end(QCowL2Meta
*m
)
456 return m
->offset
+ m
->cow_end
.offset
457 + (m
->cow_end
.nb_sectors
<< BDRV_SECTOR_BITS
);
460 // FIXME Need qcow2_ prefix to global functions
462 /* qcow2.c functions */
463 int qcow2_backing_read1(BlockDriverState
*bs
, QEMUIOVector
*qiov
,
464 int64_t sector_num
, int nb_sectors
);
466 int qcow2_mark_dirty(BlockDriverState
*bs
);
467 int qcow2_mark_corrupt(BlockDriverState
*bs
);
468 int qcow2_mark_consistent(BlockDriverState
*bs
);
469 int qcow2_update_header(BlockDriverState
*bs
);
471 /* qcow2-refcount.c functions */
472 int qcow2_refcount_init(BlockDriverState
*bs
);
473 void qcow2_refcount_close(BlockDriverState
*bs
);
475 int qcow2_update_cluster_refcount(BlockDriverState
*bs
, int64_t cluster_index
,
476 int addend
, enum qcow2_discard_type type
);
478 int64_t qcow2_alloc_clusters(BlockDriverState
*bs
, uint64_t size
);
479 int qcow2_alloc_clusters_at(BlockDriverState
*bs
, uint64_t offset
,
481 int64_t qcow2_alloc_bytes(BlockDriverState
*bs
, int size
);
482 void qcow2_free_clusters(BlockDriverState
*bs
,
483 int64_t offset
, int64_t size
,
484 enum qcow2_discard_type type
);
485 void qcow2_free_any_clusters(BlockDriverState
*bs
, uint64_t l2_entry
,
486 int nb_clusters
, enum qcow2_discard_type type
);
488 int qcow2_update_snapshot_refcount(BlockDriverState
*bs
,
489 int64_t l1_table_offset
, int l1_size
, int addend
);
491 int qcow2_check_refcounts(BlockDriverState
*bs
, BdrvCheckResult
*res
,
494 void qcow2_process_discards(BlockDriverState
*bs
, int ret
);
496 int qcow2_check_metadata_overlap(BlockDriverState
*bs
, int ign
, int64_t offset
,
498 int qcow2_pre_write_overlap_check(BlockDriverState
*bs
, int ign
, int64_t offset
,
501 /* qcow2-cluster.c functions */
502 int qcow2_grow_l1_table(BlockDriverState
*bs
, uint64_t min_size
,
504 int qcow2_write_l1_entry(BlockDriverState
*bs
, int l1_index
);
505 void qcow2_l2_cache_reset(BlockDriverState
*bs
);
506 int qcow2_decompress_cluster(BlockDriverState
*bs
, uint64_t cluster_offset
);
507 void qcow2_encrypt_sectors(BDRVQcowState
*s
, int64_t sector_num
,
508 uint8_t *out_buf
, const uint8_t *in_buf
,
509 int nb_sectors
, int enc
,
512 int qcow2_get_cluster_offset(BlockDriverState
*bs
, uint64_t offset
,
513 int *num
, uint64_t *cluster_offset
);
514 int qcow2_alloc_cluster_offset(BlockDriverState
*bs
, uint64_t offset
,
515 int *num
, uint64_t *host_offset
, QCowL2Meta
**m
);
516 uint64_t qcow2_alloc_compressed_cluster_offset(BlockDriverState
*bs
,
518 int compressed_size
);
520 int qcow2_alloc_cluster_link_l2(BlockDriverState
*bs
, QCowL2Meta
*m
);
521 int qcow2_discard_clusters(BlockDriverState
*bs
, uint64_t offset
,
522 int nb_sectors
, enum qcow2_discard_type type
);
523 int qcow2_zero_clusters(BlockDriverState
*bs
, uint64_t offset
, int nb_sectors
);
525 int qcow2_expand_zero_clusters(BlockDriverState
*bs
);
527 /* qcow2-snapshot.c functions */
528 int qcow2_snapshot_create(BlockDriverState
*bs
, QEMUSnapshotInfo
*sn_info
);
529 int qcow2_snapshot_goto(BlockDriverState
*bs
, const char *snapshot_id
);
530 int qcow2_snapshot_delete(BlockDriverState
*bs
,
531 const char *snapshot_id
,
534 int qcow2_snapshot_list(BlockDriverState
*bs
, QEMUSnapshotInfo
**psn_tab
);
535 int qcow2_snapshot_load_tmp(BlockDriverState
*bs
,
536 const char *snapshot_id
,
540 void qcow2_free_snapshots(BlockDriverState
*bs
);
541 int qcow2_read_snapshots(BlockDriverState
*bs
);
543 /* qcow2-cache.c functions */
544 Qcow2Cache
*qcow2_cache_create(BlockDriverState
*bs
, int num_tables
);
545 int qcow2_cache_destroy(BlockDriverState
* bs
, Qcow2Cache
*c
);
547 void qcow2_cache_entry_mark_dirty(Qcow2Cache
*c
, void *table
);
548 int qcow2_cache_flush(BlockDriverState
*bs
, Qcow2Cache
*c
);
549 int qcow2_cache_set_dependency(BlockDriverState
*bs
, Qcow2Cache
*c
,
550 Qcow2Cache
*dependency
);
551 void qcow2_cache_depends_on_flush(Qcow2Cache
*c
);
553 int qcow2_cache_empty(BlockDriverState
*bs
, Qcow2Cache
*c
);
555 int qcow2_cache_get(BlockDriverState
*bs
, Qcow2Cache
*c
, uint64_t offset
,
557 int qcow2_cache_get_empty(BlockDriverState
*bs
, Qcow2Cache
*c
, uint64_t offset
,
559 int qcow2_cache_put(BlockDriverState
*bs
, Qcow2Cache
*c
, void **table
);