GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux KVM nSVM: L1 guest can enter L2 with EFER.LMA=1 and LME=0, breaking KVM MMU invariants

UnscoredCVE-2026-99344Kernel, userspace & hypervisorcurated

Impact

On AMD hosts with nested virtualization enabled, VMRUN accepts the nonsensical EFER.LMA=1 with LME=0, and KVM carries that state into L2 even though its MMU assumes LMA implies LME. The fix description says the broken invariant can cause a variety of issues, especially in MMU code choosing between 4/5-level and PAE paging, and the related CVE-2026-99345 shows that this state can drive out-of-bounds accesses in the shadow MMU. A tenant guest that runs its own hypervisor reaches this path directly.

Who can reach it

A tenant with kernel privileges in an L1 guest on an AMD host where nested SVM is exposed to guests. No host authentication needed.

What to do

Install a kernel with the stable fix (commits linked) on AMD KVM hosts, then migrate or drain guests and reboot. Hosts that do not expose nested SVM (kvm_amd nested=0) are not reachable from guests. The record does not give fixed versions.

References

Related entries

All Kernel, userspace & hypervisor entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.