CVE-2021-3468 is an infinite-loop flaw (CWE-835) in avahi-daemon 0.6 to 0.8: when a local client fills the daemon's input buffer on its Unix socket and then disconnects, client_work() never handles the hang-up event, the daemon spins at full CPU and service discovery stops. NVD scores it 5.5; v0.9-rc1 fixes it.
Vulnerability summary
CVSS base score
5.5
Medium severity
EPSS score (FIRST)
0.4%
Probability of exploitation in the wild in the next 30 days
CISA KEV catalog
No
Not listed
Published
Vulnerability ID: CVE-2021-3468 · Source: NVD
Affected components
| Component | CPE | Version range |
|---|---|---|
| Avahi (avahi-daemon) | cpe:2.3:a:avahi:avahi:*:*:*:*:*:*:*:* | 0.6 up to and including 0.8 (fixed upstream in v0.9-rc1, commit 447affe; backported in Buildroot avahi 0.8 patch 0003 and Debian 0.8-5+deb11u2) |
What does this mean for affected devices?
CVE-2021-3468 is a local denial of service in the Avahi mDNS/DNS-SD daemon. According to NVD, versions 0.6 up to 0.8 mishandle the event that signals the end of a client connection on the avahi Unix socket in client_work, "allowing a local attacker to trigger an infinite loop", after which the avahi service becomes unresponsive. NVD scores it CVSS v3.1 5.5 (AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H) and classifies it as CWE-835, a loop with an unreachable exit condition.
The fix is commit 447affe, dated 2021-03-26. Its message explains the mechanism: once a client fills the input buffer, the daemon stops watching for reads, so when that client closes the socket the kernel keeps reporting a hang-up that nothing consumes. Avahi's own advisory GHSA-mrwc-c54v-3rg8 lists versions from 0.6 up to v0.9-rc1 as vulnerable and v0.9-rc1 as patched. The last final Avahi release is still 0.8 from 2020-02-18, so most products that ship Avahi carry 0.8 plus backported patches rather than the fixed tag.
That is exactly how embedded Linux build systems handle it. Buildroot's avahi package builds Avahi 0.8, applies the upstream commit as 0003-avoid-infinite-loop-in-avahi-daemon-by-handling-hup-event-in-client-work.patch and marks CVE-2021-3468 as ignored for its CVE scan because the patch is present. A firmware image built from an older Buildroot or Yocto tree without that patch reports Avahi 0.8 and is affected; one built from a current tree also reports 0.8 and is not. Our guide to Yocto and Buildroot SBOMs shows where each build system records the version and patch list, and firmware diffing and patch analysis shows how to confirm the fix landed in the shipped binary.
Distributions backported the same commit. The Debian security tracker lists fixes in stretch 0.6.32-2+deb9u1 (DLA-3047-1), buster 0.7-4+deb10u3 (DLA-3466-1) and bullseye 0.8-5+deb11u2, and bookworm 0.8-10+deb12u1 as fixed.
On a device the attacker needs a local process that can connect to /run/avahi-daemon/socket, so the risk depends on what else runs on the box: a compromised web interface, a third-party app or a container. The effect is that a printer, camera, NAS or gateway drops out of DNS-SD discovery until avahi-daemon is restarted. As of 2026-11-16 the CVE is not in the CISA KEV catalog (version 2026.11.13), and FIRST EPSS gives it 0.00426 (37th percentile) in its 2026-11-15 dataset. Our Avahi and mDNS/DNS-SD vulnerabilities guide ranks it below the network-reachable CNAME crashes such as CVE-2026-24401 and alongside the D-Bus assertions such as CVE-2023-38469.
Evidence
- Avahi 0.6 up to 0.8 mishandles client connection termination on the Unix socket in client_work, letting a local attacker trigger an infinite loop; CVSS v3.1 5.5, CWE-835. — NVD
- Avahi advisory GHSA-mrwc-c54v-3rg8 lists >= 0.6, < 0.9-rc1 as vulnerable and v0.9-rc1 as patched. — Avahi GHSA-mrwc-c54v-3rg8
- Buildroot builds Avahi 0.8 with patch 0003 (upstream commit 447affe) and ignores CVE-2021-3468 in its CVE scan. — Buildroot package/avahi/avahi.mk
- Debian fixed it in stretch 0.6.32-2+deb9u1 (DLA-3047-1), buster 0.7-4+deb10u3 (DLA-3466-1) and bullseye 0.8-5+deb11u2. — Debian Security Tracker
- Not in the CISA KEV catalog version 2026.11.13 (checked 2026-11-16). — CISA KEV
- EPSS 0.00426, 37th percentile, FIRST dataset of 2026-11-15. — FIRST EPSS

Related advisories
Sources
Frequently asked questions
Can CVE-2021-3468 be exploited over the network?
No. NVD scores it with a local attack vector (AV:L) and low privileges: the attacker needs a process on the device that can connect to the avahi-daemon Unix socket. The effect is an infinite loop that leaves the service unresponsive (per NVD).
Which Avahi version fixes CVE-2021-3468?
Upstream fixed it in commit 447affe, first tagged in v0.9-rc1; Avahi's advisory GHSA-mrwc-c54v-3rg8 lists v0.9-rc1 as patched. Because 0.8 is still the last final release, most products apply the commit as a backport instead (per the Avahi advisory).
My firmware SBOM says Avahi 0.8. Is it affected?
It depends on the patches. Buildroot builds Avahi 0.8 with patch 0003, which is the upstream fix, and Debian ships it in 0.8-5+deb11u2 and later. Check the build system's patch list or diff the binary; the version string alone does not tell you (per Buildroot avahi.mk and the Debian tracker).
Is CVE-2021-3468 being exploited?
It is not in the CISA KEV catalog (version 2026.11.13, checked 2026-11-16), and FIRST EPSS gives it 0.00426, the 37th percentile, in its 2026-11-15 dataset (per CISA KEV and FIRST EPSS).
Questions about this page?