CVE-2026-68238

What it actually does

The vulnerability is a reference leak in the Linux kernel AMD GPU driver: amdgpu_acpi_vfct_bios calls acpi_get_table to fetch the VFCT ACPI table but omits the matching acpi_put_table on all exit paths, leaving the table mapped and its validation_count incremented on every invocation.

Why it matters to the business

Sustained graphics-driver activity on AMD GPUs gradually consumes kernel memory mappings, which can degrade system stability and performance over extended uptime on affected machines.

How an attack unfolds

An attacker with local access repeatedly exercises graphics or display code paths that invoke amdgpu_acpi_vfct_bios; each call leaks a table reference until kernel resources are exhausted, producing a local denial-of-service condition.

How to fix it

Apply the upstream kernel commit that routes all exit paths through a common acpi_put_table call after the VBIOS image has been safely copied with kmemdup; rebuild and deploy the updated kernel on all affected systems.

How to confirm the fix worked

Inspect the amdgpu_acpi_vfct_bios function in the running kernel source or binary to confirm that acpi_put_table is invoked on every return path after acpi_get_table succeeds.

Typical attack path

1. Obtain local user or service account on a Linux system with an AMD GPU
2. Trigger repeated calls to amdgpu_acpi_vfct_bios via graphics operations or driver ioctls
3. Accumulate unreleased ACPI table mappings until memory pressure or instability occurs

How to detect it

Review kernel logs and ACPI subsystem traces for repeated amdgpu_acpi_vfct_bios entries; correlate with rising validation_count values or unmapped VFCT table references via kernel memory diagnostics.

What your security team should do

• Prioritize kernel updates on any AMD-GPU Linux hosts
• Monitor kernel memory and ACPI table reference counters for anomalous growth
• Restrict unprivileged local access on systems where immediate patching is delayed