CVE-2026-98466

A flaw was found in the Linux kernel's KVM (Kernel-based Virtual Machine) subsystem on ARM64 platforms. During system initialization in protected virtualization modes, the hypervisor fails to revoke host stage-2 access permissions for per-CPU hypervisor stack memory pages. A privileged local host can exploit this flaw to inspect or alter sensitive hypervisor stack data, resulting in a bypass of hypervisor isolation and potential privilege escalation.

This vulnerability is rated as an Important severity because a local attacker with elevated administrative access on the host can compromise hypervisor confidentiality and integrity, breaking the guest isolation boundary. Exploitation requires running protected KVM on ARM64 platforms, an environment specifically configured to restrict host access to guest memory. Systems operating under default configurations without protected virtualization, as well as those running on other architectures, are not exposed to this security flaw.

The CVSS score(s) listed for this vulnerability do not reflect the associated product's status, and are included for informational purposes to better understand the severity of this vulnerability.

--- CME Mitigation Guidance (2026-10-20) --- Disclaimer: The following mitigation solutions are provided as a courtesy, pending the application of an official patch. This content was generated utilizing AI tools in conjunction with data from https://cmetaxonomy.org. Customers are advised to independently evaluate and test these temporary mitigations within their own infrastructure.

The highest-value action is to revoke host stage-2 page-table mappings for per-CPU hypervisor stack memory during protected nVHE initialization: upgrade to Linux kernel 6.18.55, 7.2.9, 7.3 (or the fixed Kernel/Kernel Rt/Openshift/Ose Rhel Coreos/Kpatch builds), which walk each per-CPU hypervisor stack virtual address range using the ownership-transition walker to transfer pages out of host stage-2, closing all 2 injection vectors (host stage-2 hypervisor stack memory read, host stage-2 hypervisor stack memory modification). Until patched, this is only a concern on ARM64 Linux systems running protected KVM (pKVM / protected nVHE) mode — Non-ARM64 architectures and deployments without protected KVM mode enabled are unaffected — so Strictly limit root access and prevent privileged userspace processes from reading physical memory or interacting with debugging interfaces on the host. As a compensating measure, Enforce SELinux in enforcing mode to prevent privileged processes from accessing raw physical memory interfaces or debugging utilities capable of reading hypervisor stack pages and Enable Kernel Lockdown Mode to restrict privileged host userspace from reading or writing raw physical memory via /dev/mem and patching running kernel memory. These controls limit blast radius and contain exploitation, but they do not fix the underlying hypervisor stack memory isolation defect — apply the upstream patch to fully remediate.

Highest-impact CME: CME-301 SELinux (Enforcing Mode) (medium)