CVE-2026-98720

What it actually does

Failure to refresh the vmcs02 MSR bitmap on nested VM-Enter when runtime eVMCS controls change leaves TPR passthrough enabled after TPR virtualization is disabled, allowing L2 (or L1 by proxy) to read and write L0's TPR.

Why it matters to the business

Virtual machine isolation can be weakened in nested KVM environments, potentially exposing host-level configuration to guests and increasing risk of unauthorized access or control-plane interference.

How an attack unfolds

An attacker controlling L1 configures L2 with TPR virtualization enabled, then toggles it off for the same L2 instance; the missing MSR bitmap refresh leaves TPR passthrough active, permitting direct access to L0's TPR.

How to fix it

Update to a Linux kernel containing the fix that forces MSR bitmap refresh on runtime eVMCS control changes during nested VM-Enter; apply the patch or newer kernel release as soon as available.

How to confirm the fix worked

Confirm the kernel version includes the commit that adds the bitmap refresh logic; test nested VM-Enter sequences with toggling TPR virtualization to verify msr_bitmap02 is updated.

Typical attack path

1. L1 sets up nested virtualization for L2 with TPR virtualization initially enabled
2. L1 disables TPR virtualization for the same L2 without triggering a full VMCS refresh
3. KVM fails to update msr_bitmap02, leaving TPR passthrough
4. L2 or L1 reads/writes L0's TPR through the stale passthrough path

How to detect it

Review nested virtualization logs and KVM trace events for unexpected TPR passthrough after control toggles; correlate with guest MSR access patterns that bypass expected intercepts.

What your security team should do

• Prioritize kernel updates for all KVM hosts supporting nested virtualization
• Monitor upstream stable and distribution kernel releases for the fix
• Review nested virtualization usage and disable it where not required until patched