KVM on the x86 architecture contains a bug where the hypervisor ignores a pending Physical Vector (PV) End‑of‑Interrupt (EOI) after the virtual CPU has disabled PV EOIs. If the kernel attempts to assert that PV EOIs are enabled while a guest memory read fails, the internal check that would normally clear the pending PV EOI condition exits early, leaving the system with an unsatisfied EOI request. This leads to an invalid opcode fault and triggers a kernel BUG, resulting in an oops and a host kernel crash.

The flaw exists in the Linux kernel’s KVM module and affects all Linux distributions that ship the kernel source unchanged from the upstream repository. No specific version bounds are documented in the advisory, so any kernel where the KVM x86 code path includes the unpatched logic may be vulnerable.

The CVSS score of 7.1 reflects a high severity denial‑of‑service impact, while the EPSS score of <1% indicates a very low, though non‑zero, exploitation probability. The vulnerability is not listed in the CISA KEV catalog. Based on the description, it is inferred that an attacker would need to cause a PV EOI pending state for a disabled vCPU – for example by manipulating a guest or the hypervisor interface – to trigger the bug. This would result in a kernel panic, causing Denial of Service on the host or hypervisor. The attack requires execution within a virtual machine that can exercise this code path, and its success depends on the ability to induce the specific KVM path described.