3 A 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,
5 both for actual guest data and for image metadata.
7 Likewise, the virtual disk as seen by the guest is divided into (guest)
8 clusters of the same size.
10 All numbers in qcow2 are stored in Big Endian byte order.
15 The first cluster of a qcow2 image contains the file header:
18 QCOW magic string ("QFI\xfb")
21 Version number (valid values are 2 and 3)
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.
28 16 - 19: backing_file_size
29 Length of the backing file name in bytes. Must not be
30 longer than 1023 bytes. Undefined if the image doesn't have
34 Number of bits that are used for addressing an offset
35 within a cluster (1 << cluster_bits is the cluster size).
36 Must not be less than 9 (i.e. 512 byte clusters).
38 Note: qemu as of today has an implementation limit of 2 MB
39 as the maximum cluster size and won't be able to open images
40 with larger cluster sizes.
43 Virtual disk size in bytes
50 Number of entries in the active L1 table
52 40 - 47: l1_table_offset
53 Offset into the image file at which the active L1 table
54 starts. Must be aligned to a cluster boundary.
56 48 - 55: refcount_table_offset
57 Offset into the image file at which the refcount table
58 starts. Must be aligned to a cluster boundary.
60 56 - 59: refcount_table_clusters
61 Number of clusters that the refcount table occupies
64 Number of snapshots contained in the image
66 64 - 71: snapshots_offset
67 Offset into the image file at which the snapshot table
68 starts. Must be aligned to a cluster boundary.
70 If the version is 3 or higher, the header has the following additional fields.
71 For version 2, the values are assumed to be zero, unless specified otherwise
72 in the description of a field.
74 72 - 79: incompatible_features
75 Bitmask of incompatible features. An implementation must
76 fail to open an image if an unknown bit is set.
78 Bit 0: Dirty bit. If this bit is set then refcounts
79 may be inconsistent, make sure to scan L1/L2
80 tables to repair refcounts before accessing the
83 Bit 1: Corrupt bit. If this bit is set then any data
84 structure may be corrupt and the image must not
85 be written to (unless for regaining
88 Bits 2-63: Reserved (set to 0)
90 80 - 87: compatible_features
91 Bitmask of compatible features. An implementation can
92 safely ignore any unknown bits that are set.
94 Bit 0: Lazy refcounts bit. If this bit is set then
95 lazy refcount updates can be used. This means
96 marking the image file dirty and postponing
97 refcount metadata updates.
99 Bits 1-63: Reserved (set to 0)
101 88 - 95: autoclear_features
102 Bitmask of auto-clear features. An implementation may only
103 write to an image with unknown auto-clear features if it
104 clears the respective bits from this field first.
106 Bits 0-63: Reserved (set to 0)
108 96 - 99: refcount_order
109 Describes the width of a reference count block entry (width
110 in bits = 1 << refcount_order). For version 2 images, the
111 order is always assumed to be 4 (i.e. the width is 16 bits).
113 100 - 103: header_length
114 Length of the header structure in bytes. For version 2
115 images, the length is always assumed to be 72 bytes.
117 Directly after the image header, optional sections called header extensions can
118 be stored. Each extension has a structure like the following:
120 Byte 0 - 3: Header extension type:
121 0x00000000 - End of the header extension area
122 0xE2792ACA - Backing file format name
123 0x6803f857 - Feature name table
124 other - Unknown header extension, can be safely
127 4 - 7: Length of the header extension data
129 8 - n: Header extension data
131 n - m: Padding to round up the header extension size to the next
134 Unless stated otherwise, each header extension type shall appear at most once
137 The remaining space between the end of the header extension area and the end of
138 the first cluster can be used for the backing file name. It is not allowed to
139 store other data here, so that an implementation can safely modify the header
140 and add extensions without harming data of compatible features that it
141 doesn't support. Compatible features that need space for additional data can
142 use a header extension.
145 == Feature name table ==
147 The feature name table is an optional header extension that contains the name
148 for features used by the image. It can be used by applications that don't know
149 the respective feature (e.g. because the feature was introduced only later) to
150 display a useful error message.
152 The number of entries in the feature name table is determined by the length of
153 the header extension data. Each entry look like this:
155 Byte 0: Type of feature (select feature bitmap)
156 0: Incompatible feature
157 1: Compatible feature
160 1: Bit number within the selected feature bitmap (valid
163 2 - 47: Feature name (padded with zeros, but not necessarily null
164 terminated if it has full length)
167 == Host cluster management ==
169 qcow2 manages the allocation of host clusters by maintaining a reference count
170 for each host cluster. A refcount of 0 means that the cluster is free, 1 means
171 that it is used, and >= 2 means that it is used and any write access must
172 perform a COW (copy on write) operation.
174 The refcounts are managed in a two-level table. The first level is called
175 refcount table and has a variable size (which is stored in the header). The
176 refcount table can cover multiple clusters, however it needs to be contiguous
179 It contains pointers to the second level structures which are called refcount
180 blocks and are exactly one cluster in size.
182 Given a offset into the image file, the refcount of its cluster can be obtained
185 refcount_block_entries = (cluster_size / sizeof(uint16_t))
187 refcount_block_index = (offset / cluster_size) % refcount_block_entries
188 refcount_table_index = (offset / cluster_size) / refcount_block_entries
190 refcount_block = load_cluster(refcount_table[refcount_table_index]);
191 return refcount_block[refcount_block_index];
193 Refcount table entry:
195 Bit 0 - 8: Reserved (set to 0)
197 9 - 63: Bits 9-63 of the offset into the image file at which the
198 refcount block starts. Must be aligned to a cluster
201 If this is 0, the corresponding refcount block has not yet
202 been allocated. All refcounts managed by this refcount block
205 Refcount block entry (x = refcount_bits - 1):
207 Bit 0 - x: Reference count of the cluster. If refcount_bits implies a
208 sub-byte width, note that bit 0 means the least significant
212 == Cluster mapping ==
214 Just as for refcounts, qcow2 uses a two-level structure for the mapping of
215 guest clusters to host clusters. They are called L1 and L2 table.
217 The L1 table has a variable size (stored in the header) and may use multiple
218 clusters, however it must be contiguous in the image file. L2 tables are
219 exactly one cluster in size.
221 Given a offset into the virtual disk, the offset into the image file can be
224 l2_entries = (cluster_size / sizeof(uint64_t))
226 l2_index = (offset / cluster_size) % l2_entries
227 l1_index = (offset / cluster_size) / l2_entries
229 l2_table = load_cluster(l1_table[l1_index]);
230 cluster_offset = l2_table[l2_index];
232 return cluster_offset + (offset % cluster_size)
236 Bit 0 - 8: Reserved (set to 0)
238 9 - 55: Bits 9-55 of the offset into the image file at which the L2
239 table starts. Must be aligned to a cluster boundary. If the
240 offset is 0, the L2 table and all clusters described by this
241 L2 table are unallocated.
243 56 - 62: Reserved (set to 0)
245 63: 0 for an L2 table that is unused or requires COW, 1 if its
246 refcount is exactly one. This information is only accurate
247 in the active L1 table.
251 Bit 0 - 61: Cluster descriptor
253 62: 0 for standard clusters
254 1 for compressed clusters
256 63: 0 for a cluster that is unused or requires COW, 1 if its
257 refcount is exactly one. This information is only accurate
258 in L2 tables that are reachable from the the active L1
261 Standard Cluster Descriptor:
263 Bit 0: If set to 1, the cluster reads as all zeros. The host
264 cluster offset can be used to describe a preallocation,
265 but it won't be used for reading data from this cluster,
266 nor is data read from the backing file if the cluster is
269 With version 2, this is always 0.
271 1 - 8: Reserved (set to 0)
273 9 - 55: Bits 9-55 of host cluster offset. Must be aligned to a
274 cluster boundary. If the offset is 0, the cluster is
277 56 - 61: Reserved (set to 0)
280 Compressed Clusters Descriptor (x = 62 - (cluster_bits - 8)):
282 Bit 0 - x: Host cluster offset. This is usually _not_ aligned to a
285 x+1 - 61: Compressed size of the images in sectors of 512 bytes
287 If a cluster is unallocated, read requests shall read the data from the backing
288 file (except if bit 0 in the Standard Cluster Descriptor is set). If there is
289 no backing file or the backing file is smaller than the image, they shall read
290 zeros for all parts that are not covered by the backing file.
295 qcow2 supports internal snapshots. Their basic principle of operation is to
296 switch the active L1 table, so that a different set of host clusters are
297 exposed to the guest.
299 When creating a snapshot, the L1 table should be copied and the refcount of all
300 L2 tables and clusters reachable from this L1 table must be increased, so that
301 a write causes a COW and isn't visible in other snapshots.
303 When loading a snapshot, bit 63 of all entries in the new active L1 table and
304 all L2 tables referenced by it must be reconstructed from the refcount table
305 as it doesn't need to be accurate in inactive L1 tables.
307 A directory of all snapshots is stored in the snapshot table, a contiguous area
308 in the image file, whose starting offset and length are given by the header
309 fields snapshots_offset and nb_snapshots. The entries of the snapshot table
310 have variable length, depending on the length of ID, name and extra data.
312 Snapshot table entry:
314 Byte 0 - 7: Offset into the image file at which the L1 table for the
315 snapshot starts. Must be aligned to a cluster boundary.
317 8 - 11: Number of entries in the L1 table of the snapshots
319 12 - 13: Length of the unique ID string describing the snapshot
321 14 - 15: Length of the name of the snapshot
323 16 - 19: Time at which the snapshot was taken in seconds since the
326 20 - 23: Subsecond part of the time at which the snapshot was taken
329 24 - 31: Time that the guest was running until the snapshot was
332 32 - 35: Size of the VM state in bytes. 0 if no VM state is saved.
333 If there is VM state, it starts at the first cluster
334 described by first L1 table entry that doesn't describe a
335 regular guest cluster (i.e. VM state is stored like guest
336 disk content, except that it is stored at offsets that are
337 larger than the virtual disk presented to the guest)
339 36 - 39: Size of extra data in the table entry (used for future
340 extensions of the format)
342 variable: Extra data for future extensions. Unknown fields must be
343 ignored. Currently defined are (offset relative to snapshot
346 Byte 40 - 47: Size of the VM state in bytes. 0 if no VM
347 state is saved. If this field is present,
348 the 32-bit value in bytes 32-35 is ignored.
350 Byte 48 - 55: Virtual disk size of the snapshot in bytes
352 Version 3 images must include extra data at least up to
355 variable: Unique ID string for the snapshot (not null terminated)
357 variable: Name of the snapshot (not null terminated)
359 variable: Padding to round up the snapshot table entry size to the