xref: /openbmc/qemu/docs/interop/qcow2.txt (revision 7a1bfee6)
1== General ==
2
3A qcow2 image file is organized in units of constant size, which are called
4(host) clusters. A cluster is the unit in which all allocations are done,
5both for actual guest data and for image metadata.
6
7Likewise, the virtual disk as seen by the guest is divided into (guest)
8clusters of the same size.
9
10All numbers in qcow2 are stored in Big Endian byte order.
11
12
13== Header ==
14
15The first cluster of a qcow2 image contains the file header:
16
17    Byte  0 -  3:   magic
18                    QCOW magic string ("QFI\xfb")
19
20          4 -  7:   version
21                    Version number (valid values are 2 and 3)
22
23          8 - 15:   backing_file_offset
24                    Offset into the image file at which the backing file name
25                    is stored (NB: The string is not null terminated). 0 if the
26                    image doesn't have a backing file.
27
28                    Note: backing files are incompatible with raw external data
29                    files (auto-clear feature bit 1).
30
31         16 - 19:   backing_file_size
32                    Length of the backing file name in bytes. Must not be
33                    longer than 1023 bytes. Undefined if the image doesn't have
34                    a backing file.
35
36         20 - 23:   cluster_bits
37                    Number of bits that are used for addressing an offset
38                    within a cluster (1 << cluster_bits is the cluster size).
39                    Must not be less than 9 (i.e. 512 byte clusters).
40
41                    Note: qemu as of today has an implementation limit of 2 MB
42                    as the maximum cluster size and won't be able to open images
43                    with larger cluster sizes.
44
45         24 - 31:   size
46                    Virtual disk size in bytes.
47
48                    Note: qemu has an implementation limit of 32 MB as
49                    the maximum L1 table size.  With a 2 MB cluster
50                    size, it is unable to populate a virtual cluster
51                    beyond 2 EB (61 bits); with a 512 byte cluster
52                    size, it is unable to populate a virtual size
53                    larger than 128 GB (37 bits).  Meanwhile, L1/L2
54                    table layouts limit an image to no more than 64 PB
55                    (56 bits) of populated clusters, and an image may
56                    hit other limits first (such as a file system's
57                    maximum size).
58
59         32 - 35:   crypt_method
60                    0 for no encryption
61                    1 for AES encryption
62                    2 for LUKS encryption
63
64         36 - 39:   l1_size
65                    Number of entries in the active L1 table
66
67         40 - 47:   l1_table_offset
68                    Offset into the image file at which the active L1 table
69                    starts. Must be aligned to a cluster boundary.
70
71         48 - 55:   refcount_table_offset
72                    Offset into the image file at which the refcount table
73                    starts. Must be aligned to a cluster boundary.
74
75         56 - 59:   refcount_table_clusters
76                    Number of clusters that the refcount table occupies
77
78         60 - 63:   nb_snapshots
79                    Number of snapshots contained in the image
80
81         64 - 71:   snapshots_offset
82                    Offset into the image file at which the snapshot table
83                    starts. Must be aligned to a cluster boundary.
84
85For version 2, the header is exactly 72 bytes in length, and finishes here.
86For version 3 or higher, the header length is at least 104 bytes, including
87the next fields through header_length.
88
89         72 -  79:  incompatible_features
90                    Bitmask of incompatible features. An implementation must
91                    fail to open an image if an unknown bit is set.
92
93                    Bit 0:      Dirty bit.  If this bit is set then refcounts
94                                may be inconsistent, make sure to scan L1/L2
95                                tables to repair refcounts before accessing the
96                                image.
97
98                    Bit 1:      Corrupt bit.  If this bit is set then any data
99                                structure may be corrupt and the image must not
100                                be written to (unless for regaining
101                                consistency).
102
103                    Bit 2:      External data file bit.  If this bit is set, an
104                                external data file is used. Guest clusters are
105                                then stored in the external data file. For such
106                                images, clusters in the external data file are
107                                not refcounted. The offset field in the
108                                Standard Cluster Descriptor must match the
109                                guest offset and neither compressed clusters
110                                nor internal snapshots are supported.
111
112                                An External Data File Name header extension may
113                                be present if this bit is set.
114
115                    Bit 3:      Compression type bit.  If this bit is set,
116                                a non-default compression is used for compressed
117                                clusters. The compression_type field must be
118                                present and not zero.
119
120                    Bits 4-63:  Reserved (set to 0)
121
122         80 -  87:  compatible_features
123                    Bitmask of compatible features. An implementation can
124                    safely ignore any unknown bits that are set.
125
126                    Bit 0:      Lazy refcounts bit.  If this bit is set then
127                                lazy refcount updates can be used.  This means
128                                marking the image file dirty and postponing
129                                refcount metadata updates.
130
131                    Bits 1-63:  Reserved (set to 0)
132
133         88 -  95:  autoclear_features
134                    Bitmask of auto-clear features. An implementation may only
135                    write to an image with unknown auto-clear features if it
136                    clears the respective bits from this field first.
137
138                    Bit 0:      Bitmaps extension bit
139                                This bit indicates consistency for the bitmaps
140                                extension data.
141
142                                It is an error if this bit is set without the
143                                bitmaps extension present.
144
145                                If the bitmaps extension is present but this
146                                bit is unset, the bitmaps extension data must be
147                                considered inconsistent.
148
149                    Bit 1:      Raw external data bit
150                                If this bit is set, the external data file can
151                                be read as a consistent standalone raw image
152                                without looking at the qcow2 metadata.
153
154                                Setting this bit has a performance impact for
155                                some operations on the image (e.g. writing
156                                zeros requires writing to the data file instead
157                                of only setting the zero flag in the L2 table
158                                entry) and conflicts with backing files.
159
160                                This bit may only be set if the External Data
161                                File bit (incompatible feature bit 1) is also
162                                set.
163
164                    Bits 2-63:  Reserved (set to 0)
165
166         96 -  99:  refcount_order
167                    Describes the width of a reference count block entry (width
168                    in bits: refcount_bits = 1 << refcount_order). For version 2
169                    images, the order is always assumed to be 4
170                    (i.e. refcount_bits = 16).
171                    This value may not exceed 6 (i.e. refcount_bits = 64).
172
173        100 - 103:  header_length
174                    Length of the header structure in bytes. For version 2
175                    images, the length is always assumed to be 72 bytes.
176                    For version 3 it's at least 104 bytes and must be a multiple
177                    of 8.
178
179
180=== Additional fields (version 3 and higher) ===
181
182In general, these fields are optional and may be safely ignored by the software,
183as well as filled by zeros (which is equal to field absence), if software needs
184to set field B, but does not care about field A which precedes B. More
185formally, additional fields have the following compatibility rules:
186
1871. If the value of the additional field must not be ignored for correct
188handling of the file, it will be accompanied by a corresponding incompatible
189feature bit.
190
1912. If there are no unrecognized incompatible feature bits set, an unknown
192additional field may be safely ignored other than preserving its value when
193rewriting the image header.
194
1953. An explicit value of 0 will have the same behavior as when the field is not
196present*, if not altered by a specific incompatible bit.
197
198*. A field is considered not present when header_length is less than or equal
199to the field's offset. Also, all additional fields are not present for
200version 2.
201
202              104:  compression_type
203
204                    Defines the compression method used for compressed clusters.
205                    All compressed clusters in an image use the same compression
206                    type.
207
208                    If the incompatible bit "Compression type" is set: the field
209                    must be present and non-zero (which means non-zlib
210                    compression type). Otherwise, this field must not be present
211                    or must be zero (which means zlib).
212
213                    Available compression type values:
214                        0: zlib <https://www.zlib.net/>
215
216
217=== Header padding ===
218
219@header_length must be a multiple of 8, which means that if the end of the last
220additional field is not aligned, some padding is needed. This padding must be
221zeroed, so that if some existing (or future) additional field will fall into
222the padding, it will be interpreted accordingly to point [3.] of the previous
223paragraph, i.e.  in the same manner as when this field is not present.
224
225
226=== Header extensions ===
227
228Directly after the image header, optional sections called header extensions can
229be stored. Each extension has a structure like the following:
230
231    Byte  0 -  3:   Header extension type:
232                        0x00000000 - End of the header extension area
233                        0xE2792ACA - Backing file format name string
234                        0x6803f857 - Feature name table
235                        0x23852875 - Bitmaps extension
236                        0x0537be77 - Full disk encryption header pointer
237                        0x44415441 - External data file name string
238                        other      - Unknown header extension, can be safely
239                                     ignored
240
241          4 -  7:   Length of the header extension data
242
243          8 -  n:   Header extension data
244
245          n -  m:   Padding to round up the header extension size to the next
246                    multiple of 8.
247
248Unless stated otherwise, each header extension type shall appear at most once
249in the same image.
250
251If the image has a backing file then the backing file name should be stored in
252the remaining space between the end of the header extension area and the end of
253the first cluster. It is not allowed to store other data here, so that an
254implementation can safely modify the header and add extensions without harming
255data of compatible features that it doesn't support. Compatible features that
256need space for additional data can use a header extension.
257
258
259== String header extensions ==
260
261Some header extensions (such as the backing file format name and the external
262data file name) are just a single string. In this case, the header extension
263length is the string length and the string is not '\0' terminated. (The header
264extension padding can make it look like a string is '\0' terminated, but
265neither is padding always necessary nor is there a guarantee that zero bytes
266are used for padding.)
267
268
269== Feature name table ==
270
271The feature name table is an optional header extension that contains the name
272for features used by the image. It can be used by applications that don't know
273the respective feature (e.g. because the feature was introduced only later) to
274display a useful error message.
275
276The number of entries in the feature name table is determined by the length of
277the header extension data. Each entry look like this:
278
279    Byte       0:   Type of feature (select feature bitmap)
280                        0: Incompatible feature
281                        1: Compatible feature
282                        2: Autoclear feature
283
284               1:   Bit number within the selected feature bitmap (valid
285                    values: 0-63)
286
287          2 - 47:   Feature name (padded with zeros, but not necessarily null
288                    terminated if it has full length)
289
290
291== Bitmaps extension ==
292
293The bitmaps extension is an optional header extension. It provides the ability
294to store bitmaps related to a virtual disk. For now, there is only one bitmap
295type: the dirty tracking bitmap, which tracks virtual disk changes from some
296point in time.
297
298The data of the extension should be considered consistent only if the
299corresponding auto-clear feature bit is set, see autoclear_features above.
300
301The fields of the bitmaps extension are:
302
303    Byte  0 -  3:  nb_bitmaps
304                   The number of bitmaps contained in the image. Must be
305                   greater than or equal to 1.
306
307                   Note: Qemu currently only supports up to 65535 bitmaps per
308                   image.
309
310          4 -  7:  Reserved, must be zero.
311
312          8 - 15:  bitmap_directory_size
313                   Size of the bitmap directory in bytes. It is the cumulative
314                   size of all (nb_bitmaps) bitmap directory entries.
315
316         16 - 23:  bitmap_directory_offset
317                   Offset into the image file at which the bitmap directory
318                   starts. Must be aligned to a cluster boundary.
319
320== Full disk encryption header pointer ==
321
322The full disk encryption header must be present if, and only if, the
323'crypt_method' header requires metadata. Currently this is only true
324of the 'LUKS' crypt method. The header extension must be absent for
325other methods.
326
327This header provides the offset at which the crypt method can store
328its additional data, as well as the length of such data.
329
330    Byte  0 -  7:   Offset into the image file at which the encryption
331                    header starts in bytes. Must be aligned to a cluster
332                    boundary.
333    Byte  8 - 15:   Length of the written encryption header in bytes.
334                    Note actual space allocated in the qcow2 file may
335                    be larger than this value, since it will be rounded
336                    to the nearest multiple of the cluster size. Any
337                    unused bytes in the allocated space will be initialized
338                    to 0.
339
340For the LUKS crypt method, the encryption header works as follows.
341
342The first 592 bytes of the header clusters will contain the LUKS
343partition header. This is then followed by the key material data areas.
344The size of the key material data areas is determined by the number of
345stripes in the key slot and key size. Refer to the LUKS format
346specification ('docs/on-disk-format.pdf' in the cryptsetup source
347package) for details of the LUKS partition header format.
348
349In the LUKS partition header, the "payload-offset" field will be
350calculated as normal for the LUKS spec. ie the size of the LUKS
351header, plus key material regions, plus padding, relative to the
352start of the LUKS header. This offset value is not required to be
353qcow2 cluster aligned. Its value is currently never used in the
354context of qcow2, since the qcow2 file format itself defines where
355the real payload offset is, but none the less a valid payload offset
356should always be present.
357
358In the LUKS key slots header, the "key-material-offset" is relative
359to the start of the LUKS header clusters in the qcow2 container,
360not the start of the qcow2 file.
361
362Logically the layout looks like
363
364  +-----------------------------+
365  | QCow2 header                |
366  | QCow2 header extension X    |
367  | QCow2 header extension FDE  |
368  | QCow2 header extension ...  |
369  | QCow2 header extension Z    |
370  +-----------------------------+
371  | ....other QCow2 tables....  |
372  .                             .
373  .                             .
374  +-----------------------------+
375  | +-------------------------+ |
376  | | LUKS partition header   | |
377  | +-------------------------+ |
378  | | LUKS key material 1     | |
379  | +-------------------------+ |
380  | | LUKS key material 2     | |
381  | +-------------------------+ |
382  | | LUKS key material ...   | |
383  | +-------------------------+ |
384  | | LUKS key material 8     | |
385  | +-------------------------+ |
386  +-----------------------------+
387  | QCow2 cluster payload       |
388  .                             .
389  .                             .
390  .                             .
391  |                             |
392  +-----------------------------+
393
394== Data encryption ==
395
396When an encryption method is requested in the header, the image payload
397data must be encrypted/decrypted on every write/read. The image headers
398and metadata are never encrypted.
399
400The algorithms used for encryption vary depending on the method
401
402 - AES:
403
404   The AES cipher, in CBC mode, with 256 bit keys.
405
406   Initialization vectors generated using plain64 method, with
407   the virtual disk sector as the input tweak.
408
409   This format is no longer supported in QEMU system emulators, due
410   to a number of design flaws affecting its security. It is only
411   supported in the command line tools for the sake of back compatibility
412   and data liberation.
413
414 - LUKS:
415
416   The algorithms are specified in the LUKS header.
417
418   Initialization vectors generated using the method specified
419   in the LUKS header, with the physical disk sector as the
420   input tweak.
421
422== Host cluster management ==
423
424qcow2 manages the allocation of host clusters by maintaining a reference count
425for each host cluster. A refcount of 0 means that the cluster is free, 1 means
426that it is used, and >= 2 means that it is used and any write access must
427perform a COW (copy on write) operation.
428
429The refcounts are managed in a two-level table. The first level is called
430refcount table and has a variable size (which is stored in the header). The
431refcount table can cover multiple clusters, however it needs to be contiguous
432in the image file.
433
434It contains pointers to the second level structures which are called refcount
435blocks and are exactly one cluster in size.
436
437Although a large enough refcount table can reserve clusters past 64 PB
438(56 bits) (assuming the underlying protocol can even be sized that
439large), note that some qcow2 metadata such as L1/L2 tables must point
440to clusters prior to that point.
441
442Note: qemu has an implementation limit of 8 MB as the maximum refcount
443table size.  With a 2 MB cluster size and a default refcount_order of
4444, it is unable to reference host resources beyond 2 EB (61 bits); in
445the worst case, with a 512 cluster size and refcount_order of 6, it is
446unable to access beyond 32 GB (35 bits).
447
448Given an offset into the image file, the refcount of its cluster can be
449obtained as follows:
450
451    refcount_block_entries = (cluster_size * 8 / refcount_bits)
452
453    refcount_block_index = (offset / cluster_size) % refcount_block_entries
454    refcount_table_index = (offset / cluster_size) / refcount_block_entries
455
456    refcount_block = load_cluster(refcount_table[refcount_table_index]);
457    return refcount_block[refcount_block_index];
458
459Refcount table entry:
460
461    Bit  0 -  8:    Reserved (set to 0)
462
463         9 - 63:    Bits 9-63 of the offset into the image file at which the
464                    refcount block starts. Must be aligned to a cluster
465                    boundary.
466
467                    If this is 0, the corresponding refcount block has not yet
468                    been allocated. All refcounts managed by this refcount block
469                    are 0.
470
471Refcount block entry (x = refcount_bits - 1):
472
473    Bit  0 -  x:    Reference count of the cluster. If refcount_bits implies a
474                    sub-byte width, note that bit 0 means the least significant
475                    bit in this context.
476
477
478== Cluster mapping ==
479
480Just as for refcounts, qcow2 uses a two-level structure for the mapping of
481guest clusters to host clusters. They are called L1 and L2 table.
482
483The L1 table has a variable size (stored in the header) and may use multiple
484clusters, however it must be contiguous in the image file. L2 tables are
485exactly one cluster in size.
486
487The L1 and L2 tables have implications on the maximum virtual file
488size; for a given L1 table size, a larger cluster size is required for
489the guest to have access to more space.  Furthermore, a virtual
490cluster must currently map to a host offset below 64 PB (56 bits)
491(although this limit could be relaxed by putting reserved bits into
492use).  Additionally, as cluster size increases, the maximum host
493offset for a compressed cluster is reduced (a 2M cluster size requires
494compressed clusters to reside below 512 TB (49 bits), and this limit
495cannot be relaxed without an incompatible layout change).
496
497Given an offset into the virtual disk, the offset into the image file can be
498obtained as follows:
499
500    l2_entries = (cluster_size / sizeof(uint64_t))
501
502    l2_index = (offset / cluster_size) % l2_entries
503    l1_index = (offset / cluster_size) / l2_entries
504
505    l2_table = load_cluster(l1_table[l1_index]);
506    cluster_offset = l2_table[l2_index];
507
508    return cluster_offset + (offset % cluster_size)
509
510L1 table entry:
511
512    Bit  0 -  8:    Reserved (set to 0)
513
514         9 - 55:    Bits 9-55 of the offset into the image file at which the L2
515                    table starts. Must be aligned to a cluster boundary. If the
516                    offset is 0, the L2 table and all clusters described by this
517                    L2 table are unallocated.
518
519        56 - 62:    Reserved (set to 0)
520
521             63:    0 for an L2 table that is unused or requires COW, 1 if its
522                    refcount is exactly one. This information is only accurate
523                    in the active L1 table.
524
525L2 table entry:
526
527    Bit  0 -  61:   Cluster descriptor
528
529              62:   0 for standard clusters
530                    1 for compressed clusters
531
532              63:   0 for clusters that are unused, compressed or require COW.
533                    1 for standard clusters whose refcount is exactly one.
534                    This information is only accurate in L2 tables
535                    that are reachable from the active L1 table.
536
537                    With external data files, all guest clusters have an
538                    implicit refcount of 1 (because of the fixed host = guest
539                    mapping for guest cluster offsets), so this bit should be 1
540                    for all allocated clusters.
541
542Standard Cluster Descriptor:
543
544    Bit       0:    If set to 1, the cluster reads as all zeros. The host
545                    cluster offset can be used to describe a preallocation,
546                    but it won't be used for reading data from this cluster,
547                    nor is data read from the backing file if the cluster is
548                    unallocated.
549
550                    With version 2, this is always 0.
551
552         1 -  8:    Reserved (set to 0)
553
554         9 - 55:    Bits 9-55 of host cluster offset. Must be aligned to a
555                    cluster boundary. If the offset is 0 and bit 63 is clear,
556                    the cluster is unallocated. The offset may only be 0 with
557                    bit 63 set (indicating a host cluster offset of 0) when an
558                    external data file is used.
559
560        56 - 61:    Reserved (set to 0)
561
562
563Compressed Clusters Descriptor (x = 62 - (cluster_bits - 8)):
564
565    Bit  0 - x-1:   Host cluster offset. This is usually _not_ aligned to a
566                    cluster or sector boundary!  If cluster_bits is
567                    small enough that this field includes bits beyond
568                    55, those upper bits must be set to 0.
569
570         x - 61:    Number of additional 512-byte sectors used for the
571                    compressed data, beyond the sector containing the offset
572                    in the previous field. Some of these sectors may reside
573                    in the next contiguous host cluster.
574
575                    Note that the compressed data does not necessarily occupy
576                    all of the bytes in the final sector; rather, decompression
577                    stops when it has produced a cluster of data.
578
579                    Another compressed cluster may map to the tail of the final
580                    sector used by this compressed cluster.
581
582If a cluster is unallocated, read requests shall read the data from the backing
583file (except if bit 0 in the Standard Cluster Descriptor is set). If there is
584no backing file or the backing file is smaller than the image, they shall read
585zeros for all parts that are not covered by the backing file.
586
587
588== Snapshots ==
589
590qcow2 supports internal snapshots. Their basic principle of operation is to
591switch the active L1 table, so that a different set of host clusters are
592exposed to the guest.
593
594When creating a snapshot, the L1 table should be copied and the refcount of all
595L2 tables and clusters reachable from this L1 table must be increased, so that
596a write causes a COW and isn't visible in other snapshots.
597
598When loading a snapshot, bit 63 of all entries in the new active L1 table and
599all L2 tables referenced by it must be reconstructed from the refcount table
600as it doesn't need to be accurate in inactive L1 tables.
601
602A directory of all snapshots is stored in the snapshot table, a contiguous area
603in the image file, whose starting offset and length are given by the header
604fields snapshots_offset and nb_snapshots. The entries of the snapshot table
605have variable length, depending on the length of ID, name and extra data.
606
607Snapshot table entry:
608
609    Byte 0 -  7:    Offset into the image file at which the L1 table for the
610                    snapshot starts. Must be aligned to a cluster boundary.
611
612         8 - 11:    Number of entries in the L1 table of the snapshots
613
614        12 - 13:    Length of the unique ID string describing the snapshot
615
616        14 - 15:    Length of the name of the snapshot
617
618        16 - 19:    Time at which the snapshot was taken in seconds since the
619                    Epoch
620
621        20 - 23:    Subsecond part of the time at which the snapshot was taken
622                    in nanoseconds
623
624        24 - 31:    Time that the guest was running until the snapshot was
625                    taken in nanoseconds
626
627        32 - 35:    Size of the VM state in bytes. 0 if no VM state is saved.
628                    If there is VM state, it starts at the first cluster
629                    described by first L1 table entry that doesn't describe a
630                    regular guest cluster (i.e. VM state is stored like guest
631                    disk content, except that it is stored at offsets that are
632                    larger than the virtual disk presented to the guest)
633
634        36 - 39:    Size of extra data in the table entry (used for future
635                    extensions of the format)
636
637        variable:   Extra data for future extensions. Unknown fields must be
638                    ignored. Currently defined are (offset relative to snapshot
639                    table entry):
640
641                    Byte 40 - 47:   Size of the VM state in bytes. 0 if no VM
642                                    state is saved. If this field is present,
643                                    the 32-bit value in bytes 32-35 is ignored.
644
645                    Byte 48 - 55:   Virtual disk size of the snapshot in bytes
646
647                    Version 3 images must include extra data at least up to
648                    byte 55.
649
650        variable:   Unique ID string for the snapshot (not null terminated)
651
652        variable:   Name of the snapshot (not null terminated)
653
654        variable:   Padding to round up the snapshot table entry size to the
655                    next multiple of 8.
656
657
658== Bitmaps ==
659
660As mentioned above, the bitmaps extension provides the ability to store bitmaps
661related to a virtual disk. This section describes how these bitmaps are stored.
662
663All stored bitmaps are related to the virtual disk stored in the same image, so
664each bitmap size is equal to the virtual disk size.
665
666Each bit of the bitmap is responsible for strictly defined range of the virtual
667disk. For bit number bit_nr the corresponding range (in bytes) will be:
668
669    [bit_nr * bitmap_granularity .. (bit_nr + 1) * bitmap_granularity - 1]
670
671Granularity is a property of the concrete bitmap, see below.
672
673
674=== Bitmap directory ===
675
676Each bitmap saved in the image is described in a bitmap directory entry. The
677bitmap directory is a contiguous area in the image file, whose starting offset
678and length are given by the header extension fields bitmap_directory_offset and
679bitmap_directory_size. The entries of the bitmap directory have variable
680length, depending on the lengths of the bitmap name and extra data.
681
682Structure of a bitmap directory entry:
683
684    Byte 0 -  7:    bitmap_table_offset
685                    Offset into the image file at which the bitmap table
686                    (described below) for the bitmap starts. Must be aligned to
687                    a cluster boundary.
688
689         8 - 11:    bitmap_table_size
690                    Number of entries in the bitmap table of the bitmap.
691
692        12 - 15:    flags
693                    Bit
694                      0: in_use
695                         The bitmap was not saved correctly and may be
696                         inconsistent. Although the bitmap metadata is still
697                         well-formed from a qcow2 perspective, the metadata
698                         (such as the auto flag or bitmap size) or data
699                         contents may be outdated.
700
701                      1: auto
702                         The bitmap must reflect all changes of the virtual
703                         disk by any application that would write to this qcow2
704                         file (including writes, snapshot switching, etc.). The
705                         type of this bitmap must be 'dirty tracking bitmap'.
706
707                      2: extra_data_compatible
708                         This flags is meaningful when the extra data is
709                         unknown to the software (currently any extra data is
710                         unknown to Qemu).
711                         If it is set, the bitmap may be used as expected, extra
712                         data must be left as is.
713                         If it is not set, the bitmap must not be used, but
714                         both it and its extra data be left as is.
715
716                    Bits 3 - 31 are reserved and must be 0.
717
718             16:    type
719                    This field describes the sort of the bitmap.
720                    Values:
721                      1: Dirty tracking bitmap
722
723                    Values 0, 2 - 255 are reserved.
724
725             17:    granularity_bits
726                    Granularity bits. Valid values: 0 - 63.
727
728                    Note: Qemu currently supports only values 9 - 31.
729
730                    Granularity is calculated as
731                        granularity = 1 << granularity_bits
732
733                    A bitmap's granularity is how many bytes of the image
734                    accounts for one bit of the bitmap.
735
736        18 - 19:    name_size
737                    Size of the bitmap name. Must be non-zero.
738
739                    Note: Qemu currently doesn't support values greater than
740                    1023.
741
742        20 - 23:    extra_data_size
743                    Size of type-specific extra data.
744
745                    For now, as no extra data is defined, extra_data_size is
746                    reserved and should be zero. If it is non-zero the
747                    behavior is defined by extra_data_compatible flag.
748
749        variable:   extra_data
750                    Extra data for the bitmap, occupying extra_data_size bytes.
751                    Extra data must never contain references to clusters or in
752                    some other way allocate additional clusters.
753
754        variable:   name
755                    The name of the bitmap (not null terminated), occupying
756                    name_size bytes. Must be unique among all bitmap names
757                    within the bitmaps extension.
758
759        variable:   Padding to round up the bitmap directory entry size to the
760                    next multiple of 8. All bytes of the padding must be zero.
761
762
763=== Bitmap table ===
764
765Each bitmap is stored using a one-level structure (as opposed to two-level
766structures like for refcounts and guest clusters mapping) for the mapping of
767bitmap data to host clusters. This structure is called the bitmap table.
768
769Each bitmap table has a variable size (stored in the bitmap directory entry)
770and may use multiple clusters, however, it must be contiguous in the image
771file.
772
773Structure of a bitmap table entry:
774
775    Bit       0:    Reserved and must be zero if bits 9 - 55 are non-zero.
776                    If bits 9 - 55 are zero:
777                      0: Cluster should be read as all zeros.
778                      1: Cluster should be read as all ones.
779
780         1 -  8:    Reserved and must be zero.
781
782         9 - 55:    Bits 9 - 55 of the host cluster offset. Must be aligned to
783                    a cluster boundary. If the offset is 0, the cluster is
784                    unallocated; in that case, bit 0 determines how this
785                    cluster should be treated during reads.
786
787        56 - 63:    Reserved and must be zero.
788
789
790=== Bitmap data ===
791
792As noted above, bitmap data is stored in separate clusters, described by the
793bitmap table. Given an offset (in bytes) into the bitmap data, the offset into
794the image file can be obtained as follows:
795
796    image_offset(bitmap_data_offset) =
797        bitmap_table[bitmap_data_offset / cluster_size] +
798            (bitmap_data_offset % cluster_size)
799
800This offset is not defined if bits 9 - 55 of bitmap table entry are zero (see
801above).
802
803Given an offset byte_nr into the virtual disk and the bitmap's granularity, the
804bit offset into the image file to the corresponding bit of the bitmap can be
805calculated like this:
806
807    bit_offset(byte_nr) =
808        image_offset(byte_nr / granularity / 8) * 8 +
809            (byte_nr / granularity) % 8
810
811If the size of the bitmap data is not a multiple of the cluster size then the
812last cluster of the bitmap data contains some unused tail bits. These bits must
813be zero.
814
815
816=== Dirty tracking bitmaps ===
817
818Bitmaps with 'type' field equal to one are dirty tracking bitmaps.
819
820When the virtual disk is in use dirty tracking bitmap may be 'enabled' or
821'disabled'. While the bitmap is 'enabled', all writes to the virtual disk
822should be reflected in the bitmap. A set bit in the bitmap means that the
823corresponding range of the virtual disk (see above) was written to while the
824bitmap was 'enabled'. An unset bit means that this range was not written to.
825
826The software doesn't have to sync the bitmap in the image file with its
827representation in RAM after each write or metadata change. Flag 'in_use'
828should be set while the bitmap is not synced.
829
830In the image file the 'enabled' state is reflected by the 'auto' flag. If this
831flag is set, the software must consider the bitmap as 'enabled' and start
832tracking virtual disk changes to this bitmap from the first write to the
833virtual disk. If this flag is not set then the bitmap is disabled.
834