On this page (7 sections)
A GitHub-reviewed advisory rates a flaw in LXD (the Go package github.com/canonical/lxd) as critical. According to the GitHub advisory GHSA-fmc8-p6q7-75cc, improper validation of the backup compression algorithm setting lets a user inject extra arguments into a compressor command run on the host. The result is an arbitrary file write on the host, possibly leading to arbitrary command execution.
If you run LXD hosts, especially ones where several people or automation systems can create instances and backups, you should check your version and plan an update. Only one publisher covers this so far, and several details are not yet published, so this article separates what is confirmed from what is not.
What happened
The advisory was published on June 26, 2026 against canonical/lxd and updated on November 6, 2026. It notes it was reported from another advisory (GHSA-v6mj-8pf4-hhw4). The affected-version field is expressed as a Go pseudo-version, so the fix is identified by a commit-based version string rather than a conventional release number such as 5.x.
One thing to be aware of: the technical description in the advisory refers to Incus, a closely related project, in its explanation and examples (paths such as /var/lib/incus). The package listed as affected is canonical/lxd. Because the advisory text mixes the two names, confirm with your distribution or the LXD maintainers which release channel carries the fix for your install.
How the weakness works
This is an argument injection bug. The backup feature accepts a compression_algorithm value that names the program used to compress backup data. According to the advisory, the code splits the value into words and checks only the first word against an allowlist of compressors (bzip2, gzip, lz4, lzma, pigz, pzstd, xz, tar2sqfs and zstd). Any extra words are not rejected.
Those extra words are then passed as command-line options to the compressor, which the daemon runs on the host. An allowlisted program given attacker-chosen options can be made to do things it was never meant to do. The advisory states the outcome is an arbitrary file write on the host, which it says may lead to command execution. This write-up deliberately leaves out the advisory's proof-of-concept details.
The weakness class is common: an allowlist that validates the program name but not the arguments gives a false sense of safety. The safe pattern is to accept only a fixed set of complete, known values, or to reject any input containing extra tokens.
Who is realistically exposed
The advisory does not state the required privileges, so treat the following as reasoning rather than confirmed fact. The vulnerable input is a setting supplied through backup requests, so exposure depends on who can ask your LXD daemon to create backups. Hosts are likely most at risk when:
- The LXD API is reachable over the network, not just locally.
- Non-administrators, tenants, CI jobs or other automation hold client certificates or tokens that can create instance backups.
- The daemon runs with high privileges on the host. This is general background, not stated in the advisory, but the advisory does describe the write as happening on the host, so check how privileged your daemon is.
A single-user workstation that never exposes the API is likely in a better position, but updating is still the right call.
How to check your exposure
First find which LXD you run and how it was installed. The commands below are read-only. Run them on the LXD host.
lxd --version
lxc versionsnap list lxd
snap info lxdapt list --installed 2>/dev/null | grep -i lxdCompare what you see with the vendor's fix information. Because the advisory expresses the fixed version as a pseudo-version (0.0.0-20260630083100-c5184be310b1), a normal release number may not map to it obviously. If you cannot tell, ask your distribution or the LXD project which release contains that commit.
Next, find out who can reach the API and create backups. The commands below list the trusted client identities and the configured network listen address. Run them on the host.
lxc config trust list
lxc config get core.https_addressReview the output of the first command: every entry is an identity that can talk to the daemon, so confirm each one is still needed. The second command shows whether a network address is configured for the API. Any value there means the API may be reachable from other machines, so confirm firewall rules limit who can connect.
What to do now
- Identify every LXD host and its installed version.
- Update LXD to a build that includes the fix (0.0.0-20260630083100-c5184be310b1 or later) using your normal channel.
- Restrict network access to the LXD API to trusted management hosts only.
- Review the trust list and remove client certificates and tokens that are no longer needed.
- Limit which users and automation can create instance backups until you have patched.
- Check the host for unexpected recently written files in places such as cron directories (see below).
- Keep recent, tested backups of important instances stored off the host.
To update a snap-based install, refresh the snap. This can restart the daemon.
sudo snap refresh lxdsudo apt update
sudo apt install --only-upgrade lxdDetection and monitoring
The research does not describe indicators of compromise or in-the-wild use, so there is no confirmed signature to hunt for. Reasonable read-only checks follow from how the bug is described: an arbitrary file write on the host may drop files in locations the system runs automatically. The following command lists files in the system cron directories changed in the last 30 days; run it on the LXD host and review any recent, unfamiliar entries.
find /etc/cron.d /etc/cron.daily /etc/cron.hourly /etc/cron.weekly -type f -mtime -30 -lsAlso review the LXD daemon logs for backup requests you do not recognize, particularly from client identities that normally do not create backups. If you find a suspicious file or an unexplained command, treat the host as compromised and follow a proper incident process, such as the one in website hacked: what to do.
If the containers on an LXD host serve public websites, keep watching those sites from the outside as well. Webforta Monitor can alert you to uptime, certificate, DNS and header changes, which helps you notice side effects if a host is disturbed. It does not detect this specific flaw.
What is still unknown
- No CVE identifier or CVSS score appears in the research.
- The privileges an attacker needs are not stated.
- No exploitation in the wild is reported, but absence of reports is not proof of safety.
- The advisory mixes LXD and Incus wording, so confirm the fix channel for your install.
Sources
This article is based on the public sources below. Details can change as vendors and researchers publish more; check the original advisories before acting on version numbers.
