libslirp CVE-2026-95508

Requires the host to be configured with a small, non-default interface MTU - the description explicitly states the default MTU is not affected, which is the primary limiting factor. The attacker must control a guest that can reach libslirp's built-in DHCPv6 and TFTP services and supply a malicious DHCPv6 CLIENTID option or TFTP blksize option. This is a guest-to-host attack surface (QEMU user networking / slirp4netns rootless containers), so exploitation is gated on running libslirp with that misconfiguration; hosts on the default MTU are not exposed. AC:H reflects the non-default configuration dependency.

The Red Hat CVSS 3.1 vector (AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:H/A:H = 7.4 High) tells a nuanced story. AC:H and the description's explicit 'small interface MTU' precondition are the dominant risk-limiting signals: the default MTU is NOT affected, so exploitation is confined to non-default configurations. AV:N/PR:N reflect that any guest reaching libslirp's built-in DHCPv6/TFTP services can trigger the flaw without extra credentials, but 'network' here effectively means guest-to-host reachability of the emulated network stack, not open Internet exposure. Impact is I:H/A:H with C:N - consistent with a heap write-overflow primitive (memory corruption/code-exec write path) rather than a read/disclosure bug. No EPSS score, no CISA KEV listing, and no POC were provided, so there is no evidence of active exploitation or ease-of-automation. Net: this is a legitimately serious guest-to-host memory-corruption bug worth patching in virtualization/container hosts, but the non-default MTU gate plus AC:H make it a targeted, configuration-dependent issue rather than a mass-exploitable emergency. One caveat to verify: a guest overflowing the host (QEMU/slirp4netns) process arguably warrants S:C (scope change across the VM boundary); Red Hat scored S:U, which is more conservative.

On a virtualization or rootless-container host where libslirp user-mode networking is configured with a small (non-default) interface MTU, an attacker who controls a guest VM sends a crafted DHCPv6 CLIENTID option or an oversized TFTP blksize option to the emulated network services. The response builder overflows the reply buffer with attacker-controlled length and content, crashing the host process (DoS) or potentially achieving code execution in the host process context. No public exploit code was identified at time of analysis.