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 MIN_CLUSTER_BITS 9
63 #define MAX_CLUSTER_BITS 21
65 #define MIN_L2_CACHE_SIZE 1 /* cluster */
67 /* Must be at least 4 to cover all cases of refcount table growth */
68 #define MIN_REFCOUNT_CACHE_SIZE 4 /* clusters */
70 #define DEFAULT_L2_CACHE_BYTE_SIZE 1048576 /* bytes */
72 /* The refblock cache needs only a fourth of the L2 cache size to cover as many
74 #define DEFAULT_L2_REFCOUNT_SIZE_RATIO 4
76 #define DEFAULT_CLUSTER_SIZE 65536
79 #define QCOW2_OPT_LAZY_REFCOUNTS "lazy-refcounts"
80 #define QCOW2_OPT_DISCARD_REQUEST "pass-discard-request"
81 #define QCOW2_OPT_DISCARD_SNAPSHOT "pass-discard-snapshot"
82 #define QCOW2_OPT_DISCARD_OTHER "pass-discard-other"
83 #define QCOW2_OPT_OVERLAP "overlap-check"
84 #define QCOW2_OPT_OVERLAP_TEMPLATE "overlap-check.template"
85 #define QCOW2_OPT_OVERLAP_MAIN_HEADER "overlap-check.main-header"
86 #define QCOW2_OPT_OVERLAP_ACTIVE_L1 "overlap-check.active-l1"
87 #define QCOW2_OPT_OVERLAP_ACTIVE_L2 "overlap-check.active-l2"
88 #define QCOW2_OPT_OVERLAP_REFCOUNT_TABLE "overlap-check.refcount-table"
89 #define QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK "overlap-check.refcount-block"
90 #define QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE "overlap-check.snapshot-table"
91 #define QCOW2_OPT_OVERLAP_INACTIVE_L1 "overlap-check.inactive-l1"
92 #define QCOW2_OPT_OVERLAP_INACTIVE_L2 "overlap-check.inactive-l2"
93 #define QCOW2_OPT_CACHE_SIZE "cache-size"
94 #define QCOW2_OPT_L2_CACHE_SIZE "l2-cache-size"
95 #define QCOW2_OPT_REFCOUNT_CACHE_SIZE "refcount-cache-size"
97 typedef struct QCowHeader
{
100 uint64_t backing_file_offset
;
101 uint32_t backing_file_size
;
102 uint32_t cluster_bits
;
103 uint64_t size
; /* in bytes */
104 uint32_t crypt_method
;
105 uint32_t l1_size
; /* XXX: save number of clusters instead ? */
106 uint64_t l1_table_offset
;
107 uint64_t refcount_table_offset
;
108 uint32_t refcount_table_clusters
;
109 uint32_t nb_snapshots
;
110 uint64_t snapshots_offset
;
112 /* The following fields are only valid for version >= 3 */
113 uint64_t incompatible_features
;
114 uint64_t compatible_features
;
115 uint64_t autoclear_features
;
117 uint32_t refcount_order
;
118 uint32_t header_length
;
119 } QEMU_PACKED QCowHeader
;
121 typedef struct QEMU_PACKED QCowSnapshotHeader
{
122 /* header is 8 byte aligned */
123 uint64_t l1_table_offset
;
126 uint16_t id_str_size
;
132 uint64_t vm_clock_nsec
;
134 uint32_t vm_state_size
;
135 uint32_t extra_data_size
; /* for extension */
136 /* extra data follows */
139 } QCowSnapshotHeader
;
141 typedef struct QEMU_PACKED QCowSnapshotExtraData
{
142 uint64_t vm_state_size_large
;
144 } QCowSnapshotExtraData
;
147 typedef struct QCowSnapshot
{
148 uint64_t l1_table_offset
;
153 uint64_t vm_state_size
;
156 uint64_t vm_clock_nsec
;
160 typedef struct Qcow2Cache Qcow2Cache
;
162 typedef struct Qcow2UnknownHeaderExtension
{
165 QLIST_ENTRY(Qcow2UnknownHeaderExtension
) next
;
167 } Qcow2UnknownHeaderExtension
;
170 QCOW2_FEAT_TYPE_INCOMPATIBLE
= 0,
171 QCOW2_FEAT_TYPE_COMPATIBLE
= 1,
172 QCOW2_FEAT_TYPE_AUTOCLEAR
= 2,
175 /* Incompatible feature bits */
177 QCOW2_INCOMPAT_DIRTY_BITNR
= 0,
178 QCOW2_INCOMPAT_CORRUPT_BITNR
= 1,
179 QCOW2_INCOMPAT_DIRTY
= 1 << QCOW2_INCOMPAT_DIRTY_BITNR
,
180 QCOW2_INCOMPAT_CORRUPT
= 1 << QCOW2_INCOMPAT_CORRUPT_BITNR
,
182 QCOW2_INCOMPAT_MASK
= QCOW2_INCOMPAT_DIRTY
183 | QCOW2_INCOMPAT_CORRUPT
,
186 /* Compatible feature bits */
188 QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR
= 0,
189 QCOW2_COMPAT_LAZY_REFCOUNTS
= 1 << QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR
,
191 QCOW2_COMPAT_FEAT_MASK
= QCOW2_COMPAT_LAZY_REFCOUNTS
,
194 enum qcow2_discard_type
{
195 QCOW2_DISCARD_NEVER
= 0,
196 QCOW2_DISCARD_ALWAYS
,
197 QCOW2_DISCARD_REQUEST
,
198 QCOW2_DISCARD_SNAPSHOT
,
203 typedef struct Qcow2Feature
{
207 } QEMU_PACKED Qcow2Feature
;
209 typedef struct Qcow2DiscardRegion
{
210 BlockDriverState
*bs
;
213 QTAILQ_ENTRY(Qcow2DiscardRegion
) next
;
214 } Qcow2DiscardRegion
;
216 typedef uint64_t Qcow2GetRefcountFunc(const void *refcount_array
,
218 typedef void Qcow2SetRefcountFunc(void *refcount_array
,
219 uint64_t index
, uint64_t value
);
221 typedef struct BDRVQcowState
{
228 int l1_vm_state_index
;
229 int refcount_block_bits
;
230 int refcount_block_size
;
233 uint64_t cluster_offset_mask
;
234 uint64_t l1_table_offset
;
237 Qcow2Cache
* l2_table_cache
;
238 Qcow2Cache
* refcount_block_cache
;
240 uint8_t *cluster_cache
;
241 uint8_t *cluster_data
;
242 uint64_t cluster_cache_offset
;
243 QLIST_HEAD(QCowClusterAlloc
, QCowL2Meta
) cluster_allocs
;
245 uint64_t *refcount_table
;
246 uint64_t refcount_table_offset
;
247 uint32_t refcount_table_size
;
248 uint64_t free_cluster_index
;
249 uint64_t free_byte_offset
;
253 uint32_t crypt_method
; /* current crypt method, 0 if no key yet */
254 uint32_t crypt_method_header
;
255 AES_KEY aes_encrypt_key
;
256 AES_KEY aes_decrypt_key
;
257 uint64_t snapshots_offset
;
259 unsigned int nb_snapshots
;
260 QCowSnapshot
*snapshots
;
264 bool use_lazy_refcounts
;
267 uint64_t refcount_max
;
269 Qcow2GetRefcountFunc
*get_refcount
;
270 Qcow2SetRefcountFunc
*set_refcount
;
272 bool discard_passthrough
[QCOW2_DISCARD_MAX
];
274 int overlap_check
; /* bitmask of Qcow2MetadataOverlap values */
275 bool signaled_corruption
;
277 uint64_t incompatible_features
;
278 uint64_t compatible_features
;
279 uint64_t autoclear_features
;
281 size_t unknown_header_fields_size
;
282 void* unknown_header_fields
;
283 QLIST_HEAD(, Qcow2UnknownHeaderExtension
) unknown_header_ext
;
284 QTAILQ_HEAD (, Qcow2DiscardRegion
) discards
;
287 /* Backing file path and format as stored in the image (this is not the
288 * effective path/format, which may be the result of a runtime option
290 char *image_backing_file
;
291 char *image_backing_format
;
296 typedef struct Qcow2COWRegion
{
298 * Offset of the COW region in bytes from the start of the first cluster
299 * touched by the request.
303 /** Number of sectors to copy */
308 * Describes an in-flight (part of a) write request that writes to clusters
309 * that are not referenced in their L2 table yet.
311 typedef struct QCowL2Meta
313 /** Guest offset of the first newly allocated cluster */
316 /** Host offset of the first newly allocated cluster */
317 uint64_t alloc_offset
;
320 * Number of sectors from the start of the first allocated cluster to
321 * the end of the (possibly shortened) request
325 /** Number of newly allocated clusters */
329 * Requests that overlap with this allocation and wait to be restarted
330 * when the allocating request has completed.
332 CoQueue dependent_requests
;
335 * The COW Region between the start of the first allocated cluster and the
336 * area the guest actually writes to.
338 Qcow2COWRegion cow_start
;
341 * The COW Region between the area the guest actually writes to and the
342 * end of the last allocated cluster.
344 Qcow2COWRegion cow_end
;
346 /** Pointer to next L2Meta of the same write request */
347 struct QCowL2Meta
*next
;
349 QLIST_ENTRY(QCowL2Meta
) next_in_flight
;
353 QCOW2_CLUSTER_UNALLOCATED
,
354 QCOW2_CLUSTER_NORMAL
,
355 QCOW2_CLUSTER_COMPRESSED
,
359 typedef enum QCow2MetadataOverlap
{
360 QCOW2_OL_MAIN_HEADER_BITNR
= 0,
361 QCOW2_OL_ACTIVE_L1_BITNR
= 1,
362 QCOW2_OL_ACTIVE_L2_BITNR
= 2,
363 QCOW2_OL_REFCOUNT_TABLE_BITNR
= 3,
364 QCOW2_OL_REFCOUNT_BLOCK_BITNR
= 4,
365 QCOW2_OL_SNAPSHOT_TABLE_BITNR
= 5,
366 QCOW2_OL_INACTIVE_L1_BITNR
= 6,
367 QCOW2_OL_INACTIVE_L2_BITNR
= 7,
369 QCOW2_OL_MAX_BITNR
= 8,
372 QCOW2_OL_MAIN_HEADER
= (1 << QCOW2_OL_MAIN_HEADER_BITNR
),
373 QCOW2_OL_ACTIVE_L1
= (1 << QCOW2_OL_ACTIVE_L1_BITNR
),
374 QCOW2_OL_ACTIVE_L2
= (1 << QCOW2_OL_ACTIVE_L2_BITNR
),
375 QCOW2_OL_REFCOUNT_TABLE
= (1 << QCOW2_OL_REFCOUNT_TABLE_BITNR
),
376 QCOW2_OL_REFCOUNT_BLOCK
= (1 << QCOW2_OL_REFCOUNT_BLOCK_BITNR
),
377 QCOW2_OL_SNAPSHOT_TABLE
= (1 << QCOW2_OL_SNAPSHOT_TABLE_BITNR
),
378 QCOW2_OL_INACTIVE_L1
= (1 << QCOW2_OL_INACTIVE_L1_BITNR
),
379 /* NOTE: Checking overlaps with inactive L2 tables will result in bdrv
381 QCOW2_OL_INACTIVE_L2
= (1 << QCOW2_OL_INACTIVE_L2_BITNR
),
382 } QCow2MetadataOverlap
;
384 /* Perform all overlap checks which can be done in constant time */
385 #define QCOW2_OL_CONSTANT \
386 (QCOW2_OL_MAIN_HEADER | QCOW2_OL_ACTIVE_L1 | QCOW2_OL_REFCOUNT_TABLE | \
387 QCOW2_OL_SNAPSHOT_TABLE)
389 /* Perform all overlap checks which don't require disk access */
390 #define QCOW2_OL_CACHED \
391 (QCOW2_OL_CONSTANT | QCOW2_OL_ACTIVE_L2 | QCOW2_OL_REFCOUNT_BLOCK | \
392 QCOW2_OL_INACTIVE_L1)
394 /* Perform all overlap checks */
395 #define QCOW2_OL_ALL \
396 (QCOW2_OL_CACHED | QCOW2_OL_INACTIVE_L2)
398 #define L1E_OFFSET_MASK 0x00fffffffffffe00ULL
399 #define L2E_OFFSET_MASK 0x00fffffffffffe00ULL
400 #define L2E_COMPRESSED_OFFSET_SIZE_MASK 0x3fffffffffffffffULL
402 #define REFT_OFFSET_MASK 0xfffffffffffffe00ULL
404 static inline int64_t start_of_cluster(BDRVQcowState
*s
, int64_t offset
)
406 return offset
& ~(s
->cluster_size
- 1);
409 static inline int64_t offset_into_cluster(BDRVQcowState
*s
, int64_t offset
)
411 return offset
& (s
->cluster_size
- 1);
414 static inline int size_to_clusters(BDRVQcowState
*s
, int64_t size
)
416 return (size
+ (s
->cluster_size
- 1)) >> s
->cluster_bits
;
419 static inline int64_t size_to_l1(BDRVQcowState
*s
, int64_t size
)
421 int shift
= s
->cluster_bits
+ s
->l2_bits
;
422 return (size
+ (1ULL << shift
) - 1) >> shift
;
425 static inline int offset_to_l2_index(BDRVQcowState
*s
, int64_t offset
)
427 return (offset
>> s
->cluster_bits
) & (s
->l2_size
- 1);
430 static inline int64_t align_offset(int64_t offset
, int n
)
432 offset
= (offset
+ n
- 1) & ~(n
- 1);
436 static inline int64_t qcow2_vm_state_offset(BDRVQcowState
*s
)
438 return (int64_t)s
->l1_vm_state_index
<< (s
->cluster_bits
+ s
->l2_bits
);
441 static inline uint64_t qcow2_max_refcount_clusters(BDRVQcowState
*s
)
443 return QCOW_MAX_REFTABLE_SIZE
>> s
->cluster_bits
;
446 static inline int qcow2_get_cluster_type(uint64_t l2_entry
)
448 if (l2_entry
& QCOW_OFLAG_COMPRESSED
) {
449 return QCOW2_CLUSTER_COMPRESSED
;
450 } else if (l2_entry
& QCOW_OFLAG_ZERO
) {
451 return QCOW2_CLUSTER_ZERO
;
452 } else if (!(l2_entry
& L2E_OFFSET_MASK
)) {
453 return QCOW2_CLUSTER_UNALLOCATED
;
455 return QCOW2_CLUSTER_NORMAL
;
459 /* Check whether refcounts are eager or lazy */
460 static inline bool qcow2_need_accurate_refcounts(BDRVQcowState
*s
)
462 return !(s
->incompatible_features
& QCOW2_INCOMPAT_DIRTY
);
465 static inline uint64_t l2meta_cow_start(QCowL2Meta
*m
)
467 return m
->offset
+ m
->cow_start
.offset
;
470 static inline uint64_t l2meta_cow_end(QCowL2Meta
*m
)
472 return m
->offset
+ m
->cow_end
.offset
473 + (m
->cow_end
.nb_sectors
<< BDRV_SECTOR_BITS
);
476 static inline uint64_t refcount_diff(uint64_t r1
, uint64_t r2
)
478 return r1
> r2
? r1
- r2
: r2
- r1
;
481 // FIXME Need qcow2_ prefix to global functions
483 /* qcow2.c functions */
484 int qcow2_backing_read1(BlockDriverState
*bs
, QEMUIOVector
*qiov
,
485 int64_t sector_num
, int nb_sectors
);
487 int qcow2_mark_dirty(BlockDriverState
*bs
);
488 int qcow2_mark_corrupt(BlockDriverState
*bs
);
489 int qcow2_mark_consistent(BlockDriverState
*bs
);
490 int qcow2_update_header(BlockDriverState
*bs
);
492 void qcow2_signal_corruption(BlockDriverState
*bs
, bool fatal
, int64_t offset
,
493 int64_t size
, const char *message_format
, ...)
496 /* qcow2-refcount.c functions */
497 int qcow2_refcount_init(BlockDriverState
*bs
);
498 void qcow2_refcount_close(BlockDriverState
*bs
);
500 int qcow2_get_refcount(BlockDriverState
*bs
, int64_t cluster_index
,
503 int qcow2_update_cluster_refcount(BlockDriverState
*bs
, int64_t cluster_index
,
504 uint64_t addend
, bool decrease
,
505 enum qcow2_discard_type type
);
507 int64_t qcow2_alloc_clusters(BlockDriverState
*bs
, uint64_t size
);
508 int qcow2_alloc_clusters_at(BlockDriverState
*bs
, uint64_t offset
,
510 int64_t qcow2_alloc_bytes(BlockDriverState
*bs
, int size
);
511 void qcow2_free_clusters(BlockDriverState
*bs
,
512 int64_t offset
, int64_t size
,
513 enum qcow2_discard_type type
);
514 void qcow2_free_any_clusters(BlockDriverState
*bs
, uint64_t l2_entry
,
515 int nb_clusters
, enum qcow2_discard_type type
);
517 int qcow2_update_snapshot_refcount(BlockDriverState
*bs
,
518 int64_t l1_table_offset
, int l1_size
, int addend
);
520 int qcow2_check_refcounts(BlockDriverState
*bs
, BdrvCheckResult
*res
,
523 void qcow2_process_discards(BlockDriverState
*bs
, int ret
);
525 int qcow2_check_metadata_overlap(BlockDriverState
*bs
, int ign
, int64_t offset
,
527 int qcow2_pre_write_overlap_check(BlockDriverState
*bs
, int ign
, int64_t offset
,
530 /* qcow2-cluster.c functions */
531 int qcow2_grow_l1_table(BlockDriverState
*bs
, uint64_t min_size
,
533 int qcow2_write_l1_entry(BlockDriverState
*bs
, int l1_index
);
534 void qcow2_l2_cache_reset(BlockDriverState
*bs
);
535 int qcow2_decompress_cluster(BlockDriverState
*bs
, uint64_t cluster_offset
);
536 void qcow2_encrypt_sectors(BDRVQcowState
*s
, int64_t sector_num
,
537 uint8_t *out_buf
, const uint8_t *in_buf
,
538 int nb_sectors
, int enc
,
541 int qcow2_get_cluster_offset(BlockDriverState
*bs
, uint64_t offset
,
542 int *num
, uint64_t *cluster_offset
);
543 int qcow2_alloc_cluster_offset(BlockDriverState
*bs
, uint64_t offset
,
544 int *num
, uint64_t *host_offset
, QCowL2Meta
**m
);
545 uint64_t qcow2_alloc_compressed_cluster_offset(BlockDriverState
*bs
,
547 int compressed_size
);
549 int qcow2_alloc_cluster_link_l2(BlockDriverState
*bs
, QCowL2Meta
*m
);
550 int qcow2_discard_clusters(BlockDriverState
*bs
, uint64_t offset
,
551 int nb_sectors
, enum qcow2_discard_type type
, bool full_discard
);
552 int qcow2_zero_clusters(BlockDriverState
*bs
, uint64_t offset
, int nb_sectors
);
554 int qcow2_expand_zero_clusters(BlockDriverState
*bs
,
555 BlockDriverAmendStatusCB
*status_cb
);
557 /* qcow2-snapshot.c functions */
558 int qcow2_snapshot_create(BlockDriverState
*bs
, QEMUSnapshotInfo
*sn_info
);
559 int qcow2_snapshot_goto(BlockDriverState
*bs
, const char *snapshot_id
);
560 int qcow2_snapshot_delete(BlockDriverState
*bs
,
561 const char *snapshot_id
,
564 int qcow2_snapshot_list(BlockDriverState
*bs
, QEMUSnapshotInfo
**psn_tab
);
565 int qcow2_snapshot_load_tmp(BlockDriverState
*bs
,
566 const char *snapshot_id
,
570 void qcow2_free_snapshots(BlockDriverState
*bs
);
571 int qcow2_read_snapshots(BlockDriverState
*bs
);
573 /* qcow2-cache.c functions */
574 Qcow2Cache
*qcow2_cache_create(BlockDriverState
*bs
, int num_tables
);
575 int qcow2_cache_destroy(BlockDriverState
* bs
, Qcow2Cache
*c
);
577 void qcow2_cache_entry_mark_dirty(BlockDriverState
*bs
, Qcow2Cache
*c
,
579 int qcow2_cache_flush(BlockDriverState
*bs
, Qcow2Cache
*c
);
580 int qcow2_cache_set_dependency(BlockDriverState
*bs
, Qcow2Cache
*c
,
581 Qcow2Cache
*dependency
);
582 void qcow2_cache_depends_on_flush(Qcow2Cache
*c
);
584 int qcow2_cache_empty(BlockDriverState
*bs
, Qcow2Cache
*c
);
586 int qcow2_cache_get(BlockDriverState
*bs
, Qcow2Cache
*c
, uint64_t offset
,
588 int qcow2_cache_get_empty(BlockDriverState
*bs
, Qcow2Cache
*c
, uint64_t offset
,
590 void qcow2_cache_put(BlockDriverState
*bs
, Qcow2Cache
*c
, void **table
);