Summary
- CVE-2026-89775 affects the Linux kernel’s KVM virtualisation subsystem on ARM64 processors.
- The flaw exposes freed host memory to a guest VM on systems where nested virtualisation is active.
- A guest can exploit the vulnerability to read and write host kernel memory.
- The discovering researcher assessed that guest-to-host escape and arbitrary code execution on the host are achievable.
- No corroborating sources have independently confirmed patch availability or active exploitation at time of publication.
What the vulnerability is
A use-after-free condition in the Linux kernel’s KVM virtualisation code for ARM64 processors has been assigned CVE-2026-89775. The flaw causes a freed region of host memory to remain accessible to a guest virtual machine. Because the memory is no longer under the kernel’s active management when the guest accesses it, the usual boundary controls that separate guest and host address spaces do not apply.
Scope and conditions
The vulnerability is present specifically on ARM64 hosts running the Linux kernel’s KVM hypervisor with nested virtualisation enabled. Nested virtualisation — the ability to run a hypervisor inside a virtual machine — is not a default configuration on most production deployments, but it is commonly enabled in cloud development environments, CI/CD pipelines, and certain managed Kubernetes node configurations where workloads themselves require virtualisation capabilities.
What an attacker can do
According to the researcher who discovered the flaw, successful exploitation allows a malicious or compromised guest to both read and write arbitrary host kernel memory. The researcher further assessed that this capability is sufficient to escape the guest environment entirely and execute code directly on the underlying host machine. The practical consequence is that tenant isolation — the foundational security guarantee of any multi-tenant virtualised environment — can be broken by a guest with sufficient access to trigger the vulnerable code path.
What remains unknown
The primary source does not confirm whether a patch has been merged into the Linux kernel mainline or stable trees, nor whether any Linux distribution has issued an advisory or updated package at the time of publication. There is no reported evidence of active exploitation in the wild. The precise trigger mechanism and the level of privilege required inside the guest to reach the vulnerable code path are not detailed in the available source material.
Context for security teams
ARM64-based infrastructure has grown considerably in enterprise and cloud environments over the past several years, driven in part by cost and energy efficiency advantages. AWS Graviton, Ampere Altra, and similar platforms are now common in production. Any organisation running ARM64 Linux hosts with KVM and nested virtualisation should treat this vulnerability as warranting prompt review of their exposure, irrespective of whether a patch is immediately to hand.
Why it matters
For CISOs overseeing cloud, hybrid, or on-premises infrastructure built on ARM64 Linux with KVM, this vulnerability strikes at the core isolation guarantee between workloads. A guest-to-host escape means that a compromised or malicious workload can, in principle, pivot to the host and from there reach every other tenant or workload on the same physical machine. Multi-tenant environments — including internal private clouds and shared CI/CD infrastructure — carry the highest exposure. The nested virtualisation requirement narrows the affected population, but organisations should not assume they are unaffected without actively verifying their hypervisor configuration.
What to do now
- Audit all ARM64 Linux hosts running KVM to determine whether nested virtualisation is enabled; disable it on any host where it is not operationally required.
- Monitor the upstream Linux kernel mailing lists and your Linux distribution’s security advisories for patch availability specific to CVE-2026-89775.
- Apply patches as soon as they are available from your distribution vendor, prioritising production hosts that support multi-tenant or shared workloads.
- Review cloud and infrastructure provider advisories if you rely on managed ARM64 compute services, as provider-side patching timelines and exposure may differ from self-managed environments.
