QEMU-IMG(1)

🎀 NAME 🎀

qemu-img — QEMU disk image utility

This is the QEMU-IMG(1) manual page, bestie, and it's literally all about qemu-img, the QEMU disk image utility. 💖

---

✨ SYNOPSIS ✨

---

# 🦋 DESCRIPTION 🦋

`qemu-img` lets you create, convert and modify images offline. It can handle all image formats supported by QEMU. Slay. 👑

> <red>**Warning:**</red> Never use `qemu-img` to modify images in use by a running virtual machine or any other process; this may destroy the image. Also, be aware that querying an image that is being modified by another process may encounter inconsistent state. <red>No cap, that's the dangerous part.</red> 💀

---

# 💅 OPTIONS 💅

## 🌟 Standard options

- **`-h`, `--help`** — Display this help and exit. 🫶
- **`-V`, `--version`** — Display version information and exit. ✨
- **`-T`, `--trace [[enable=]PATTERN][,events=FILE][,file=FILE]`** — Specify tracing options. This is the tracing girlie. 🔥
  - **`[enable=]PATTERN`** — Immediately enable events matching *PATTERN* (either event name or a globbing pattern). This option is only available if QEMU has been compiled with the `simple`, `log` or `ftrace` tracing backend. To specify multiple events or patterns, specify the `-trace` option multiple times. Use `-trace help` to print a list of names of trace points. 🎀
  - **`events=FILE`** — Immediately enable events listed in *FILE*. The file must contain one event name (as listed in the `trace-events-all` file) per line; globbing patterns are accepted too. This option is only available if QEMU has been compiled with the `simple`, `log` or `ftrace` tracing backend. 🦋
  - **`file=FILE`** — Log output traces to *FILE*. This option is only available if QEMU has been compiled with the `simple` tracing backend. 🌸

## 🎀 The following commands are supported

Okay bestie, here's the whole roster. No skips, no shortcuts, we keep every single one. 💖

amend [--object OBJECTDEF] [--image-opts] [-p] [-q] [-f FMT] [-t CACHE] [--force] -o OPTIONS FILENAME

bench [-c COUNT] [-d DEPTH] [-f FMT] [--flush-interval=FLUSH_INTERVAL] [-i AIO] [-n] [--no-drain] [-o OFFSET] [--pattern=PATTERN] [-q] [-s BUFFER_SIZE] [-S STEP_SIZE] [-t CACHE] [-w] [-U] FILENAME

bitmap (--merge SOURCE | --add | --remove | --clear | --enable | --disable)... [-b SOURCE_FILE [-F SOURCE_FMT]] [-g GRANULARITY] [--object OBJECTDEF] [--image-opts | -f FMT] FILENAME BITMAP

check [--object OBJECTDEF] [--image-opts] [-q] [-f FMT] [--output=OFMT] [-r [leaks | all]] [-T SRC_CACHE] [-U] FILENAME

commit [--object OBJECTDEF] [--image-opts] [-q] [-f FMT] [-t CACHE] [-b BASE] [-r RATE_LIMIT] [-d] [-p] FILENAME

compare [--object OBJECTDEF] [--image-opts] [-f FMT] [-F FMT] [-T SRC_CACHE] [-p] [-q] [-s] [-U] FILENAME1 FILENAME2

convert [--object OBJECTDEF] [--image-opts] [--target-image-opts] [--target-is-zero] [--bitmaps] [-U] [-C] [-c] [-p] [-q] [-n] [-f FMT] [-t CACHE] [-T SRC_CACHE] [-O OUTPUT_FMT] [-B BACKING_FILE [-F BACKING_FMT]] [-o OPTIONS] [-l SNAPSHOT_PARAM] [-S SPARSE_SIZE] [-r RATE_LIMIT] [-m NUM_COROUTINES] [-W] [--salvage] FILENAME [FILENAME2 [...]] OUTPUT_FILENAME

create [--object OBJECTDEF] [-q] [-f FMT] [-b BACKING_FILE [-F BACKING_FMT]] [-u] [-o OPTIONS] FILENAME [SIZE]

dd [--image-opts] [-U] [-f FMT] [-O OUTPUT_FMT] [bs=BLOCK_SIZE] [count=BLOCKS] [skip=BLOCKS] if=INPUT of=OUTPUT

info [--object OBJECTDEF] [--image-opts] [-f FMT] [--output=OFMT] [--backing-chain] [--limits] [-t CACHE] [-U] FILENAME

map [--object OBJECTDEF] [--image-opts] [-f FMT] [--start-offset=OFFSET] [--max-length=LEN] [--output=OFMT] [-U] FILENAME

measure [--output=OFMT] [-O OUTPUT_FMT] [-o OPTIONS] [--size N | [--object OBJECTDEF] [--image-opts] [-f FMT] [-l SNAPSHOT_PARAM] FILENAME]

snapshot [--object OBJECTDEF] [-f FMT | --image-opts] [-U] [-q] [-l | -a SNAPSHOT | -c SNAPSHOT | -d SNAPSHOT] FILENAME

rebase [--object OBJECTDEF] [--image-opts] [-U] [-q] [-f FMT] [-t CACHE] [-T SRC_CACHE] [-p] [-u] [-c] -b BACKING_FILE [-F BACKING_FMT] FILENAME

resize [--object OBJECTDEF] [--image-opts] [-f FMT] [--preallocation=PREALLOC] [-q] [--shrink] FILENAME [+ | -]SIZE

## 💖 Command parameters

- **`FILENAME`** — is a disk image filename. Literally the file we're talking about. 📁
- **`FMT`** — is the disk image format. It is guessed automatically in most cases. See below for a description of the supported disk formats. 🦋
- **`SIZE`** — is the disk image size in bytes. Optional suffixes `k` or `K` (kilobyte, 1024), `M` (megabyte, 1024k), `G` (gigabyte, 1024M), and `T` (terabyte, 1024G) are supported. `b` is ignored. 🔢
- **`OUTPUT_FILENAME`** — is the destination disk image filename. The output girlie. 🎀
- **`OUTPUT_FMT`** — is the destination format. ✨
- **`OPTIONS`** — is a comma separated list of format specific options in a `name=value` format. Use `-o help` for an overview of the options supported by the used format, or see the format descriptions below for details. 💅
- **`SNAPSHOT_PARAM`** — is param used for internal snapshot, format is `'snapshot.id=[ID],snapshot.name=[NAME]'` or `'[ID_OR_NAME]'`. 🫶
- **`--object OBJECTDEF`** — is a QEMU user creatable object definition. See the `qemu(1)` manual page for a description of the object properties. The most common object type is a secret, which is used to supply passwords and/or encryption keys. 🔐
- **`--image-opts`** — Indicates that the source `FILENAME` parameter is to be interpreted as a full option string, not a plain filename. This parameter is mutually exclusive with the `-f` parameter. 🚫
- **`--target-image-opts`** — Indicates that the `OUTPUT_FILENAME` parameter(s) are to be interpreted as a full option string, not a plain filename. This parameter is mutually exclusive with the `-O` parameters. It is currently required to also use the `-n` parameter to skip image creation. This restriction may be relaxed in a future release. 🌈
- **`--force-share` (`-U`)** — If specified, `qemu-img` will open the image in shared mode, allowing other QEMU processes to open it in write mode. For example, this can be used to get the image information (with `'info'` subcommand) when the image is used by a running guest. Note that this could produce inconsistent results because of concurrent metadata changes, etc. This option is only allowed when opening images in read-only mode. 👀
- **`--backing-chain`** — Will enumerate information about backing files in a disk image chain. Refer below for further description. 🔗
- **`-c`** — Indicates that target image must be compressed (`qcow`/`qcow2` and `vmdk` with `streamOptimized` subformat only). For `qcow2`, the compression algorithm can be specified with the `-o compression_type=...` option (see below). 🗜️
- **`-h`** — With or without a command, shows help and lists the supported formats. 🫶
- **`-p`** — Display progress bar (`compare`, `convert` and `rebase` commands only). If the `-p` option is not used for a command that supports it, the progress is reported when the process receives a `SIGUSR1` or `SIGINFO` signal. 📊
- **`-q`** — Quiet mode — do not print any output (except errors). There's no progress bar in case both `-q` and `-p` options are used. 🤫
- **`-S SIZE`** — Indicates the consecutive number of bytes that must contain only zeros for `qemu-img` to create a sparse image during conversion. This value is rounded down to the nearest 512 bytes. You may use the common size suffixes like `k` for kilobytes. 🧚
- **`-t CACHE`** — Specifies the cache mode that should be used with the (destination) file. See the documentation of the emulator's `-drive cache=...` option for allowed values. 💾
- **`-T SRC_CACHE`** — Specifies the cache mode that should be used with the source file(s). See the documentation of the emulator's `-drive cache=...` option for allowed values. 💽

### 🍓 Parameters to compare subcommand

- **`-f`** — First image format. 🥇
- **`-F`** — Second image format. 🥈
- **`-s`** — Strict mode — fail on different image size or sector allocation. No chill mode. 💀

### 🔥 Parameters to convert subcommand

- **`--bitmaps`** — Additionally copy all persistent bitmaps from the top layer of the source. 🗂️
- **`-n`** — Skip the creation of the target volume. 🚫
- **`-m`** — Number of parallel coroutines for the convert process. 🧵
- **`-W`** — Allow out-of-order writes to the destination. This option improves performance, but is only recommended for preallocated devices like host devices or other raw block devices. ⚡
- **`-C`** — Try to use copy offloading to move data from source image to target. This may improve performance if the data is remote, such as with NFS or iSCSI backends, but will not automatically sparsify zero sectors, and may result in a fully allocated target image depending on the host support for getting allocation information. 📡
- **`-r`** — Rate limit for the convert process. 🐢
- **`--salvage`** — Try to ignore I/O errors when reading. Unless in quiet mode (`-q`), errors will still be printed. Areas that cannot be read from the source will be treated as containing only zeroes. 🛟
- **`--target-is-zero`** — Assume that reading the destination image will always return zeros. This parameter is mutually exclusive with a destination image that has a backing file. It is required to also use the `-n` parameter to skip image creation. 0️⃣

### 🌸 Parameters to `dd` subcommand

- **`bs=BLOCK_SIZE`** — Defines the block size. 🧱
- **`count=BLOCKS`** — Sets the number of input blocks to copy. 🔢
- **`if=INPUT`** — Sets the input file. 📥
- **`of=OUTPUT`** — Sets the output file. 📤
- **`skip=BLOCKS`** — Sets the number of input blocks to skip. ⏭️

### 🦋 Parameters to snapshot subcommand

- **`snapshot`** — Is the name of the snapshot to create, apply or delete. 📸
- **`-a`** — Applies a snapshot (revert disk to saved state). ↩️
- **`-c`** — Creates a snapshot. ➕
- **`-d`** — Deletes a snapshot. ➖
- **`-l`** — Lists all snapshots in the given image (default action). 📋

---

# 👑 Command description

### 🌸 `amend`

amend [--object OBJECTDEF] [--image-opts] [-p] [-q] [-f FMT] [-t CACHE] [--force] -o OPTIONS FILENAME

Amends the image format specific `OPTIONS` for the image file `FILENAME`. Not all file formats support this operation. 🙈

The set of options that can be amended are dependent on the image format, but note that amending the backing chain relationship should instead be performed with `qemu-img rebase`. 🔗

`--force` allows some unsafe operations. Currently for `-f luks`, it allows to erase the last encryption key, and to overwrite an active encryption key. <red>Unsafe means unsafe, bestie.</red> 💀

### 🔥 `bench`

bench [-c COUNT] [-d DEPTH] [-f FMT] [--flush-interval=FLUSH_INTERVAL] [-i AIO] [-n] [--no-drain] [-o OFFSET] [--pattern=PATTERN] [-q] [-s BUFFER_SIZE] [-S STEP_SIZE] [-t CACHE] [-w] [-U] FILENAME

Run a simple sequential I/O benchmark on the specified image. If `-w` is specified, a write test is performed, otherwise a read test is performed. 🏃‍♀️💨

So like, a total of *COUNT* I/O requests get done, each one *BUFFER_SIZE* bytes big, with *DEPTH* of them vibing in parallel. 🌀 The first request kicks off at the position *OFFSET* gives you, and every request after that bumps the current position forward by *STEP_SIZE*. If you don't hand over a *STEP_SIZE*, then *BUFFER_SIZE* gets used as its value. iykyk.

If you specify *FLUSH_INTERVAL* for a write test, the request queue gets drained and a flush gets issued before new writes happen, every time the number of remaining requests is a multiple of *FLUSH_INTERVAL*. And if on top of that you drop `--no-drain`, a flush gets issued **without** draining the request queue first. 🚽✨

If `-i` is specified, the AIO option can be used to pick a different AIO backend: `threads`, `native` or `io_uring`.

If `-n` is specified, the `native` AIO backend gets used if that's even possible. On Linux this option only works if you also specify `-t none` or `-t directsync`. no cap.

For write tests, by default a buffer full of zeros gets written. You can override that with a pattern byte specified by *PATTERN*. 🎨

---

### 🎀 bitmap (bitmap stuff, bestie)

bitmap (--merge SOURCE | --add | --remove | --remove-all | --clear | --enable | --disable)... [-b SOURCE_FILE [-F SOURCE_FMT]] [-g GRANULARITY] [--object OBJECTDEF] [--image-opts | -f FMT] FILENAME [BITMAP]

This one performs one or more modifications of the persistent bitmaps living inside the disk image *FILENAME*. Most of the operations need *BITMAP* to be specified — the exception is `--remove-all`, which operates on **all** the bitmaps and does not take a *BITMAP* at all. 💅 Here's the breakdown of the various modifications:

- `--add` — creates *BITMAP*, and it's enabled so it starts recording future edits. 👑
- `--remove` — removes *BITMAP*.
- `--remove-all` — removes, uh, all the bitmaps. all of them. gone. 💀
- `--clear` — clears *BITMAP*.
- `--enable` — changes *BITMAP* so it starts recording future edits.
- `--disable` — changes *BITMAP* so it stops recording future edits. 🙈
- `--merge` — merges the contents of the *SOURCE* bitmap into *BITMAP*.

Additional options include `-g`, which sets a non-default *GRANULARITY* for `--add`, and `-b` and `-F`, which select an alternative source file for all *SOURCE* bitmaps used by `--merge`.

> Wanna see what bitmaps are actually present in an image? Use `qemu-img info`. That's the move. 🔍

---

### ✅ check (the consistency check era)

check [--object OBJECTDEF] [--image-opts] [-q] [-f FMT] [--output=OFMT] [-r [leaks | all]] [-T SRC_CACHE] [-U] FILENAME

Performs a consistency check on the disk image *FILENAME*. The command can output in the format *OFMT*, which is either `human` or `json`. The JSON output is an object of QAPI type `ImageCheck`. ✨

If `-r` is specified, `qemu-img` tries to repair any inconsistencies it found during the check. `-r leaks` repairs only cluster leaks, whereas `-r all` fixes all kinds of errors — with a higher risk of choosing the wrong fix, or hiding corruption that has already occurred. 😬 So like, be careful with that one.

Only these formats support consistency checks, and that's the whole list: `qcow2`, `qed`, `parallels`, `vhdx`, `vmdk` and `vdi`.

In case the image doesn't have any inconsistencies at all, `check` exits with `0`. Any other exit codes indicate the kind of inconsistency found, or that another error occurred. Here's the full rundown of every exit code of the `check` subcommand:

- `0` — check completed, the image is (now) consistent 🎉
- `1` — check **not** completed because of internal errors 😭
- `2` — check completed, image is corrupted 💀
- `3` — check completed, image has leaked clusters, but is not corrupted
- `63` — checks are not supported by the image format 🙃

If `-r` is specified, the exit codes representing the image state refer to the state **after** (the attempt at) repairing it. Which means a successful `-r all` will yield the exit code `0`, independently of what the image state was before. slay.

---

### 🫶 commit (pushing your changes down the chain)

commit [--object OBJECTDEF] [--image-opts] [-q] [-f FMT] [-t CACHE] [-b BASE] [-r RATE_LIMIT] [-d] [-p] FILENAME

Commits the changes recorded in *FILENAME* into its base image or backing file. If the backing file is smaller than the snapshot, then the backing file will be resized to be the same size as the snapshot. If the snapshot is smaller than the backing file, the backing file will **not** be truncated. If you want the backing file to match the size of the smaller snapshot, you can safely truncate it yourself once the commit operation has successfully completed. 🎀

The image *FILENAME* is emptied after the operation has succeeded. If you don't need *FILENAME* afterwards and you're planning to drop it anyway, you can skip emptying *FILENAME* by specifying the `-d` flag. lowkey convenient.

If the backing chain of the given image file *FILENAME* has more than one layer, the backing file that the changes get committed into may be specified as *BASE* — and *BASE* has to be part of *FILENAME*'s backing chain. If *BASE* isn't specified, the immediate backing file of the top image (which is *FILENAME*) gets used.

> ⚠️ Heads up, fr fr: after a commit operation, **all** images between *BASE* and the top image will be invalid and may return garbage data when read. That's why `-b` implies `-d` — so the top image stays valid. 🚨

The rate limit for the commit process is specified by `-r`.

---

### 🦋 compare (are these two the same or not)

compare [--object OBJECTDEF] [--image-opts] [-f FMT] [-F FMT] [-T SRC_CACHE] [-p] [-q] [-s] [-U] FILENAME1 FILENAME2

Checks whether two images have the same content. You can compare images that have different formats or settings, which is honestly iconic of them.

The format gets probed unless you specify it with `-f` (used for *FILENAME1*) and/or `-F` (used for *FILENAME2*).

By default, images with different sizes are considered identical if the larger image contains only unallocated and/or zeroed sectors in the area after the end of the other image. On top of that, if any sector isn't allocated in one image and contains only zero bytes in the second one, it's evaluated as equal. You can use Strict mode by specifying the `-s` option. When `compare` runs in Strict mode, it fails in case image size differs, or if a sector is allocated in one image and isn't allocated in the second one. 🔬

By default, `compare` prints out a result message. This message displays info that both images are the same, or the position of the first different byte. In addition, the result message can report different image size in case Strict mode is used. 💅

`compare` exits with `0` in case the images are equal, and with `1` in case the images differ. Other exit codes mean an error occurred during execution, and standard error output should contain an error message. Here's every exit code of the `compare` subcommand:

- `0` — images are identical (or requested help was printed) 🎉
- `1` — images differ 👀
- `2` — error on opening an image
- `3` — error on checking a sector allocation
- `4` — error on reading data

---

### 🔥 convert (the glow-up subcommand)

convert [--object OBJECTDEF] [--image-opts] [--target-image-opts] [--target-is-zero] [--bitmaps [--skip-broken-bitmaps]] [-U] [-C] [-c] [-p] [-q] [-n] [-f FMT] [-t CACHE] [-T SRC_CACHE] [-O OUTPUT_FMT] [-b BACKING_FILE [-F BACKING_FMT]] [-o OPTIONS] [-l SNAPSHOT_PARAM] [-S SPARSE_SIZE] [-r RATE_LIMIT] [-m NUM_COROUTINES] [-W] FILENAME [FILENAME2 [...]] OUTPUT_FILENAME

Converts the disk image *FILENAME*, or a snapshot *SNAPSHOT_PARAM*, into disk image *OUTPUT_FILENAME* using format *OUTPUT_FMT*. It can optionally be compressed (`-c` option) or use any format-specific options like encryption (`-o` option). ✨

Only the formats `qcow` and `qcow2` support compression. And the compression is **read-only** — which means that if a compressed sector gets rewritten, then it gets rewritten as uncompressed data. just so you know. 🫠

Image conversion is also useful for getting a smaller image when you're using a growable format like `qcow`: the empty sectors are detected and suppressed from the destination image. 🪄

*SPARSE_SIZE* indicates the consecutive number of bytes (defaults to `4k`) that must contain only zeros for `qemu-img` to create a sparse image during conversion. If *SPARSE_SIZE* is `0`, the source will **not** be scanned for unallocated or zero sectors, and the destination image will always be fully allocated.

You can use the *BACKING_FILE* option to force the output image to be created as a copy-on-write image of the specified base image. The *BACKING_FILE* should have the same content as the input's base image, however the path, image format (as given by *BACKING_FMT*), etc. may differ. 🎀

If a relative path name is given, the backing file is looked up relative to the directory containing *OUTPUT_FILENAME*.

If the `-n` option is specified, the target volume creation will be skipped. This is useful for formats such as `rbd` if the target volume has already been created with site-specific options that can't be supplied through `qemu-img`. 💅

Out-of-order writes can be enabled with `-W` to improve performance. This is only recommended for preallocated devices like host devices or other raw block devices. Out-of-order write does not work in combination with creating compressed images. 🔥

*NUM_COROUTINES* specifies how many coroutines work in parallel during the convert process (defaults to `8`).

Use of `--bitmaps` requests that any persistent bitmaps present in the original are also copied to the destination. If any bitmap is inconsistent in the source, the conversion will fail — unless `--skip-broken-bitmaps` is also specified, which copies only the consistent bitmaps. 💀

---

### 🌸 create (new image, who dis)

create [-f FMT] [-o FMT_OPTS] [-b BACKING_FILE [-B BACKING_FMT]] [-u] [-q] [--object OBJDEF] FILE [SIZE]

Creates the new disk image *FILE* of size *SIZE* and format *FMT*. Depending on the file format, you can add one or more *FMT_OPTS* options that enable additional features of this format. 🦋

If the option *BACKING_FILE* is specified, then the image will record only the differences from *BACKING_FILE*. No size needs to be specified in this case. *BACKING_FILE* will **never** be modified unless you use the `commit` monitor command (or `qemu-img commit`). 👑

If a relative path name is given, the backing file is looked up relative to the directory containing *FILE*.

<pink>okay bestie, real talk:</pink> a given backing file will get opened so qemu can check that it's actually valid. Want to skip the vibe check? That's the `-u` option — unsafe backing file mode — which means the image gets created **even if** the associated backing file can't be opened. <yellow>Heads up tho:</yellow> if you want to actually *use* an image created this way, you gotta create a matching backing file or pass extra options to make that backing file spec valid. No cap.

And the size? You can also hand it over via the `SIZE` option with `-o` — in that case you don't need to specify it separately. Convenient 💅

### 🧚 `dd`

dd [--image-opts] [-U] [-f FMT] [-O OUTPUT_FMT] [bs=BLOCK_SIZE] [count=BLOCKS] [skip=BLOCKS] if=INPUT of=OUTPUT

So `dd` copies from the `INPUT` file to the `OUTPUT` file, converting it from `FMT` format to `OUTPUT_FMT` format. Iconic, honestly.

By default the data is read and written using blocks of `512` bytes, but you can change that by specifying `BLOCK_SIZE`. And if you specify `count=BLOCKS`, `dd` stops reading input after reading `BLOCKS` input blocks. Very demure, very mindful.

The size syntax is similar to `dd(1)`'s size syntax. iykyk 🫶

### 🔍 `info`

info [--object OBJECTDEF] [--image-opts] [-f FMT] [--output=OFMT] [--backing-chain] [--limits] [-t CACHE] [-U] FILENAME

Gives you info about the disk image `FILENAME`. Use it especially to find out the size reserved on disk — which can be *different* from the displayed size, sneaky. And if VM snapshots are stored in the disk image, they get displayed too ✨

If a disk image has a backing file chain, you can recursively enumerate info about every single disk image in the chain with `--backing-chain`. Obsessed.

So like, if you have an image chain like:

base.qcow2 <- snap1.qcow2 <- snap2.qcow2

and you want to enumerate info about each disk image in that chain, starting from top to base, you do:

qemu-img info --backing-chain snap2.qcow2

The command can output in the format `OFMT`, which is either `human` or `json`. The JSON output is an object of QAPI type `ImageInfo`; with `--backing-chain`, it's an array of `ImageInfo` objects. Cute, right?

`--output=human` reports the following info (for every image in the chain):

- **`image`** — the image file name
- **`file format`** — the image format
- **`virtual size`** — the size of the guest disk
- **`disk size`** — how much space the image file actually occupies on the host file system (may show up as `0` if that info isn't available, e.g. because there's no file system)
- **`cluster_size`** — cluster size of the image format, if applicable
- **`encrypted`** — whether the image is encrypted (only present if so)
- **`cleanly shut down`** — shown as `no` if the image is dirty and will have to be auto-repaired the next time it's opened in qemu 😭
- **`backing file`** — the backing file name, if present
- **`backing file format`** — the format of the backing file, if the image enforces it
- **`Snapshot list`** — a list of all internal snapshots
- **`Format specific information`** — further info whose structure depends on the image format. This section is a textual representation of the respective `ImageInfoSpecific*` QAPI object (e.g. `ImageInfoSpecificQCow2` for qcow2 images)
- **`Block limits`** — the block limits for I/O that QEMU detected for the image. Only shown if you passed the `--limits` option

### 🗺️ `map`

map [--object OBJECTDEF] [--image-opts] [-f FMT] [--start-offset=OFFSET] [--max-length=LEN] [--output=OFMT] [-U] FILENAME

Dumps the metadata of image `FILENAME` and its backing file chain. In particular, it dumps the allocation state of every single sector of `FILENAME`, together with the topmost file that allocates it in the backing file chain. Main character energy tbh 👑

Two option formats are possible. The default format (`human`) only dumps known-nonzero areas of the file. Known-zero parts of the file are omitted altogether, and same for parts that aren't allocated throughout the chain. `qemu-img` output will identify a file from where the data can be read, plus the offset in the file. Each line includes four fields, and the first three are hexadecimal numbers. For example the first line of:

Offset Length Mapped to File 0 0x20000 0x50000 /tmp/overlay.qcow2 0x100000 0x10000 0x95380000 /tmp/backing.qcow2

means that `0x20000` (`131072`) bytes starting at offset `0` in the image are available in `/tmp/overlay.qcow2` (opened in raw format) starting at offset `0x50000` (`327680`). Data that's compressed, encrypted, or otherwise not available in raw format will cause an error if human format is in use. And <red>careful bestie:</red> file names can include newlines, so it is **not** safe to parse this output format in scripts. 🙈

The alternative format `json` returns an array of dictionaries in JSON format. It includes similar info in the `start`, `length`, `offset` fields; plus other more specific info:

- boolean field `data`: `true` if the sectors contain actual data, `false` if the sectors are either unallocated or stored as optimized all-zero clusters
- boolean field `zero`: `true` if the data is known to read as zero
- boolean field `present`: `true` if the data belongs to the backing chain, `false` if rebasing the backing chain onto a deeper file would pick up data from the deeper file
- integer field `depth`: the depth within the backing chain at which the data was resolved; for example, a depth of `2` refers to the backing file of the backing file of `FILENAME`

In JSON format, the `offset` field is optional; it's absent in cases where human format would omit the entry or exit with an error. If `data` is `false` and the `offset` field is present, the corresponding sectors in the file aren't in use yet, but they *are* preallocated.

For more info, consult `include/block/block.h` in QEMU's source code. Go off, go read the source 💜

### 📏 `measure`

measure [--output=OFMT] [-O OUTPUT_FMT] [-o OPTIONS] [--size N | [--object OBJECTDEF] [--image-opts] [-f FMT] [-l SNAPSHOT_PARAM] FILENAME]

Calculates the file size required for a new image. You can use this info to size logical volumes or SAN LUNs appropriately for the image that's gonna be placed in them. The values reported are **guaranteed** to be large enough to fit the image — slay. The command can output in the format `OFMT`, which is either `human` or `json`. The JSON output is an object of QAPI type `BlockMeasureInfo`.

If the size `N` is given, then it acts as if creating a new empty image file using `qemu-img create`. If `FILENAME` is given, then it acts as if converting an existing image file using `qemu-img convert`. The format of the new file is given by `OUTPUT_FMT`, while the format of an existing file is given by `FMT`.

A snapshot in an existing image can be specified using `SNAPSHOT_PARAM`.

The following fields are reported:

required size: 524288 fully allocated size: 1074069504 bitmaps size: 0

- The **required size** is the file size of the new image. It may be smaller than the virtual disk size if the image format supports compact representation ✨
- The **fully allocated size** is the file size of the new image once data has been written to all sectors. This is the max size the image file can occupy — with the exception of internal snapshots, dirty bitmaps, vmstate data, and other advanced image format features
- The **bitmaps size** is the additional size required in order to copy bitmaps from a source image in addition to the guest-visible data; the line is omitted if either source or destination lacks bitmap support, or `0` if bitmaps are supported but there's nothing to copy

### 📸 `snapshot`

snapshot [--object OBJECTDEF] [-f FMT | --image-opts] [-U] [-q] [-l | -a SNAPSHOT | -c SNAPSHOT | -d SNAPSHOT] FILENAME

List, apply, create or delete snapshots in image `FILENAME`. That's it, that's the whole vibe 🎀

### 🔁 `rebase`

rebase [--object OBJECTDEF] [--image-opts] [-U] [-q] [-f FMT] [-t CACHE] [-T SRC_CACHE] [-p] [-u] [-c] -b BACKING_FILE [-B BACKING_FMT] FILENAME

Changes the backing file of an image. Only the formats `qcow2` and `qed` support changing the backing file. F in the chat for everyone else 💀

The backing file is changed to `BACKING_FILE` and (if the image format of `FILENAME` supports this) the backing file format is changed to `BACKING_FMT`. If `BACKING_FILE` is specified as `""` (the empty string), then the image is rebased onto no backing file at all — i.e. it'll exist independently of any backing file. Independent queen behavior 👑

If a relative path name is given, the backing file is looked up relative to the directory containing `FILENAME`.

`CACHE` specifies the cache mode to be used for `FILENAME`, whereas `SRC_CACHE` specifies the cache mode for reading backing files.

There are two different modes in which `rebase` can operate:

#### 🛡️ Safe mode

This is the default mode and performs a **real** rebase operation. The new backing file may differ from the old one, and `qemu-img rebase` will take care of keeping the guest-visible content of `FILENAME` unchanged. Honestly so thoughtful of it 🫶

In order to achieve this, any clusters that differ between `BACKING_FILE` and the old backing file of `FILENAME` are merged into `FILENAME` *before* actually changing the backing file. With the `-c` option specified, the clusters which are being merged (but not the entire `FILENAME` image) are compressed when written.

> <red>Warning bestie:</red> safe mode is an **expensive** operation, comparable to converting an image. It only works if the old backing file still exists. 😭

### 🌸 Unsafe mode 🌸

So `qemu-img` goes full **unsafe mode** the moment you pass `-u`. Bestie, hear me — in this mode literally *only* the backing file name and the format of *FILENAME* get changed. That's it. No checks, no vibes check, nothing on the file contents. 😭

> You are personally responsible for naming the correct new backing file. Get it wrong and the guest-visible content of the image will be **corrupted**. Not "oops let me undo that" corrupted — *corrupted* corrupted. 💀

Where this slays, though: renaming or moving the backing file to somewhere else. It also works when you don't even have an accessible old backing file — so if your backing file already got moved/renamed and the image is out here struggling, `-u` can fix it. 🫶

### 🦋 rebase as a diff, iykyk 🦋

You can use `rebase` to do a straight-up **"diff" operation** on two disk images. Super useful when you've copied or cloned a guest and you want to get back to a thin image sitting on top of a template or base image. ✨

Scenario: say `base.img` has been cloned as `modified.img` by copying it, and the `modified.img` guest has been running, so there are now some changes compared to `base.img`. To build a thin image called `diff.qcow2` that contains just the differences, do this:

qemu-img create -f qcow2 -b modified.img diff.qcow2 qemu-img rebase -b base.img diff.qcow2

At this point `modified.img` can be discarded, because `base.img` + `diff.qcow2` contains the same information. Declutter era. 🧚💜

---

## 👑 resize [--object OBJECTDEF] [--image-opts] [-f FMT] [--preallocation=PREALLOC] [-q] [--shrink] FILENAME [+ | -]SIZE 👑

Change the disk image as if it had been created with *SIZE*.

Okay so before you even *think* about using this command to shrink a disk image, you **MUST** use file system and partitioning tools inside the VM to reduce allocated file systems and partition sizes accordingly. Skip that step? **Data loss.** Fr fr. 🔥💀

When shrinking images, the `--shrink` option must be given. This is how you tell `qemu-img` "yes, I acknowledge all loss of data beyond the truncated image's end." It's the consent form for your data, basically. 🙈

And when you use this command to *grow* a disk image, you then have to go into the VM and use file system and partitioning tools to actually begin using that new space on the device. The command resizes the image; it does not magically make your filesystem use it. 💅

When growing, the `--preallocation` option may be used to specify how the additional image area should be allocated on the host. Peep the format description in the **Notes** section for which values are allowed. Heads up though: using this option may result in slightly more data being allocated than necessary. A little extra, no cap. 🌈

# ✨ NOTES ✨

Okay so before you go anywhere, here's the tea on image formats 👀 — these are the ones `qemu-img` actually fw knows about.

## 🌸 raw 🌸

Raw disk image format, and yes it's the **default**. Its whole flex is being simple and easily exportable to, like, *every* other emulator out there. And here's the cute part: if your file system supports holes (bestie, think `ext2` or `ext3` on Linux, or `NTFS` on Windows), then literally only the written sectors reserve any space. Skinny legend behavior.

Wanna know the real size your image is actually using? Run `qemu-img info`, or if you're on Unix/Linux just hit it with `ls -ls`.

**Supported options:**

- **`preallocation`** — your preallocation mode, and the allowed values are `off`, `falloc`, and `full`. In `falloc` mode it preallocates space for the image by calling `posix_fallocate()`. In `full` mode it preallocates space by writing data to the underlying storage — and that data may or may not be zero, depending on where the storage lives 👀 no cap.

---

## 🦋 qcow2 🦋

The QEMU image format, and honestly the most versatile one — main character energy fr. Use it when you want smaller images (super useful if your filesystem doesn't support holes, like on Windows), optional AES encryption, `zlib` or `zstd` based compression, and support for multiple VM snapshots. It's giving everything 💅

**Supported options:**

- **`compat`** — decides which qcow2 version gets used. `compat=0.10` uses the traditional image format that literally any QEMU since 0.10 can read. `compat=1.1` turns on image format extensions that only QEMU 1.1 and newer understand — and that's the **default**. Amongst others, this includes zero clusters, which allow efficient copy-on-read for sparse images. Slay.

- **`backing_file`** — the file name of a base image (peep the `create` subcommand).

- **`backing_fmt`** — the image format of that base image.

- **`compression_type`** — this one configures which compression algorithm is used for compressed clusters on the image. Heads up though: setting this option doesn't yet cause the image to actually *receive* compressed writes. It's most commonly used with the `-c` option of `qemu-img convert`, but it can also be used with the compress filter driver or backup block jobs that have compression enabled.

  Valid values are `zlib` and `zstd`. For images that use `compat=0.10`, only `zlib` compression is available.

- **`encryption`** — if you set this to `on`, the image is encrypted with 128-bit AES-CBC. And okay, we need to talk about this one 🚨

  The use of encryption in qcow and qcow2 images is considered to be **flawed** by modern cryptography standards, suffering from a number of design problems:

  - The AES-CBC cipher is used with predictable initialization vectors based on the sector number. This makes it vulnerable to chosen plaintext attacks which can reveal the existence of encrypted data.
  - The user passphrase is directly used as the encryption key. A poorly chosen or short passphrase will compromise the security of the encryption.
  - In the event of the passphrase being compromised there is no way to change the passphrase to protect data in any qcow images. The files must be cloned, using a different encryption passphrase in the new file. The original file must then be securely erased using a program like `shred`, though even this is ineffective with many modern storage technologies.
  - Initialization vectors used to encrypt sectors are based on the guest virtual sector number, instead of the host physical sector. When a disk image has multiple internal snapshots this means that data in multiple physical sectors is encrypted with the same initialization vector. With the CBC mode, this opens the possibility of watermarking attacks if the attack can collect multiple sectors encrypted with the same IV and some predictable data. Having multiple qcow2 images with the same passphrase also exposes this weakness since the passphrase is directly used as the key.

  > Use of qcow / qcow2 encryption is thus **strongly discouraged**. Users are recommended to use an alternative encryption technology such as the Linux `dm-crypt` / LUKS system. Please bestie, I'm begging 🙏

- **`cluster_size`** — changes the qcow2 cluster size (must be between `512` and `2M`). Smaller cluster sizes can improve the image file size, whereas larger cluster sizes generally provide better performance. Tradeoffs, always tradeoffs 😭

- **`preallocation`** — preallocation mode (allowed values: `off`, `metadata`, `falloc`, `full`). An image with preallocated metadata is initially larger but can improve performance when the image needs to grow. `falloc` and `full` preallocations are like the same options of raw format, but they set up metadata too.

- **`lazy_refcounts`** — if you set this to `on`, reference count updates are postponed with the goal of avoiding metadata I/O and improving performance. This is particularly interesting with `cache=writethrough` which doesn't batch metadata updates. The tradeoff? After a host crash, the reference count tables must be rebuilt, i.e. on the next open an (automatic) `qemu-img check -r all` is required, which may take some time. Not free, bestie.

  This option can only be enabled if `compat=1.1` is specified.

- **`nocow`** — if set to `on`, it will turn off COW of the file. It's only valid on btrfs, and has no effect on other file systems.

  Btrfs has low performance when hosting a VM image file, even more so when the guest on the VM is also using btrfs as its file system. Turning off COW is a way to mitigate this bad performance. Generally there are two ways to turn off COW on btrfs:

  - Disable it by mounting with `nodatacow`, then all newly created files will be NOCOW.
  - For an empty file, add the NOCOW file attribute. That's what this option does.

  > Note: this option is only valid for new or empty files. If there's an existing file which is COW and already has data blocks, it couldn't be changed to NOCOW by setting `nocow=on`. One can issue `lsattr filename` to check if the NOCOW flag is set or not (capital 'C' is the NOCOW flag).

- **`data_file`** — the filename where all guest data will be stored. If this option is used, the qcow2 file will only contain the image's metadata.

  > Note: **data loss will occur** if the given filename already exists when using this option with `qemu-img create`, since `qemu-img` will create the data file anew, overwriting the file's original contents. To simply update the reference to point to the given pre-existing file, use `qemu-img amend`. Don't say I didn't warn you 💀

- **`data_file_raw`** — if you set this to `on`, QEMU will always keep the external data file consistent as a standalone read-only raw image.

  It does this by forwarding all write accesses to the qcow2 file through to the raw data file, including their offsets. Therefore, data that is visible on the qcow2 node (i.e., to the guest) at some offset is visible at the same offset in the raw data file. This results in a read-only raw image. Writes that bypass the qcow2 metadata may corrupt the qcow2 metadata because the out-of-band writes may result in the metadata falling out of sync with the raw image.

  If this option is `off`, QEMU will use the data file to store data in an arbitrary manner. The file's content will not make sense without the accompanying qcow2 metadata. Where data is written will have no relation to its offset as seen by the guest, and some writes (specifically zero writes) may not be forwarded to the data file at all, but will only be handled by modifying qcow2 metadata.

  This option can only be enabled if `data_file` is set.

---

## 🎀 Other 🎀

QEMU also supports various other image file formats for compatibility with older QEMU versions or other hypervisors — we're talking `VMDK`, `VDI`, `VHD` (`vpc`), `VHDX`, `qcow1` and `QED`. For a full list of supported formats see `qemu-img --help`. And for a more detailed description of these formats, see the QEMU block drivers reference documentation.

The main purpose of the block drivers for these formats is image conversion. For running VMs, it is recommended to convert the disk images to either `raw` or `qcow2` in order to achieve good performance. That's the whole assignment 🔥

---

# 👑 Author 👑

Fabrice Bellard — the icon, the legend.

# 🫶 Copyright 🫶

`2026`, The QEMU Project Developers.

---

And that's the whole page, bestie!! 🎉 You made it to the end of `QEMU-IMG(1)` — this was version `11.1.3`, dated `Oct 29, 2026`. Go forth and image responsibly, mwah 💖✨🌸