An inconsistency between the guest’s extended feature flags (XFD) and the XSAVE state bits (XSTATE_BV) can cause the Linux kernel to execute XRSTOR with a stale mask. When the kernel restores the FPU state, the processor raises a #NM fault, which escalates to a panic. The result is an abrupt loss of kernel availability, which can be triggered by a malicious guest that writes to the MSR controlling XFD or by manipulating the guest XSAVE contents through KVM_SET_XSAVE. The flaw does not provide attackers with privilege escalation or data exposure; it simply provides a reliable path to crash the host or other guests running on the same hypervisor.

Linux kernels starting with v5.17 and all 6.19 release candidates (rc1 through rc8) are affected. The issue manifests in any environment that uses the KVM virtualization layer and allows a guest to write to the IA32_XFD register or supply arbitrary XSAVE state via ioctl calls.

With a CVSS score of 5.5 and an EPSS below 1 percent, the vulnerability is considered moderate in severity but has a very low probability of active exploitation. It is not listed in the CISA KEV catalog. Because the attack vector requires a guest that can execute privileged MSR writes or inject XSAVE data, the risk is restricted to virtualized servers where the host trusts the guest or provides insufficient isolation. Successful exploitation results in a system-wide crash rather than loss of data or elevated privileges.