A bug in the Linux kernel’s KVM nested‑SVM (nSVM) path causes an invalid‑opcode exception, or #UD, whenever a Level 2 guest issues a VMMCALL that the host hypervisor is not configured to intercept and the call is not one of the supported Hyper‑V hypercalls. Instead of forwarding the request to the host, KVM raises the #UD, aborting the guest execution. The resulting crash or stalling of the Level 2 virtual machine constitutes a denial of service that can be triggered by any malicious or misbehaving guest when nested virtualization is enabled.

The issue affects Linux hosts running the kernel with KVM and the nested‑SVM capability enabled. Any kernel version that predates the commit adding the patch is vulnerable; administrators should refer to the provided kernel commit logs to confirm whether their running kernel contains the fix. No explicit version range is listed, so the vulnerability applies to all affected configurations prior to the fix.

The CVSS score is 7.9 while the EPSS score indicates a very low probability of exploitation (less than 1%). It is not listed in the CISA KEV catalog. The most likely attack vector is a host hypervisor granting a malicious Level 2 guest the ability to issue arbitrary VMMCALLs; this requires control of a nested virtualization environment. Because the vulnerability leads only to a guest crash rather than privilege escalation or data exfiltration, the risk level is moderate for environments that rely heavily on nested VMs, with a limited exploitation likelihood.