CVE-2023-52812

What it actually does

Out-of-bounds array access in AMD GPU PCIe parameter update code when pcie_table->num_of_link_levels evaluates to zero in SR-IOV environments, resulting in memory corruption via index calculation of num_of_levels-1.

Why it matters to the business

Kernel instability, crashes, or exploitable memory corruption on Linux systems using AMD GPUs with SR-IOV virtualization, disrupting virtual machine hosting, GPU passthrough workloads, and requiring unplanned downtime or reboots.

How an attack unfolds

An attacker with local access on an SR-IOV-enabled Linux system using AMD GPUs triggers the vulnerable drm/amd PCIe parameter update path; the zero link level count leads to an invalid array index, enabling memory corruption that can be chained for privilege escalation or denial of service. Listed threat actor campaigns reference similar Linux kernel out-of-bounds write patterns but no confirmed active exploitation of this specific issue is indicated.

How to fix it

Upgrade Linux kernel to version 6.1.119 or later; apply commit 2f2d48b6247ae3001f83c98730b3cce475cb2927 in affected stable git trees. Reboot after update and validate SR-IOV functionality.

How to confirm the fix worked

Confirm running kernel version meets or exceeds 6.1.119; inspect drm/amd source for the added num_of_link_levels bounds check before the array access; test SR-IOV initialization on AMD hardware without triggering panics.

Typical attack path

1. Gain local user access on target Linux system configured with SR-IOV and AMD GPU
2. Trigger or wait for PCIe parameter update routine in drm/amd driver
3. Force num_of_link_levels to zero condition
4. Leverage resulting out-of-bounds memory access for corruption or control flow hijack

How to detect it

Review kernel logs (dmesg, /var/log/kern.log) for AMD GPU driver oops, array index errors, or PCIe parameter update failures during SR-IOV initialization; correlate with process activity invoking drm/amd ioctls.

What your security team should do

• Apply kernel updates to 6.1.119+ on all AMD GPU hosts using SR-IOV within 30 days
• Audit virtualization hosts for AMD SR-IOV configurations and prioritize patching
• Enable kernel address space layout randomization and stack canaries where supported
• Monitor changelogs for backports to older LTS kernels still in use