CVE-2026-89775 in the Linux kernel’s KVM virtualization code for ARM64 processors exposes freed host memory to guest virtual machines, allowing guest VMs to read and write host kernel memory and potentially escape to the host system.
The vulnerability affects Linux hosts with ARM64 processors running nested virtualization. A researcher disclosed the flaw on September 22, 2026, stating it can be used to escape the guest environment and execute code on the host machine. Linux kernel maintainers have been notified, and patches are expected in upcoming releases.
CVE-2026-89775 Affects ARM64 KVM Hosts with Nested Virtualization Enabled
The flaw resides in the KVM (Kernel-based Virtual Machine) implementation for ARM64 processors in the Linux kernel. KVM provides hardware-accelerated virtualization, allowing Linux systems to run guest operating systems as virtual machines.
Nested virtualization allows a guest VM to run its own virtual machines, creating a VM-inside-a-VM configuration. This capability is used in cloud environments where tenants need to run their own hypervisors, in development and testing scenarios requiring virtualization inside VMs, and in security research involving layered virtualization architectures.
Guest VMs Gain Read and Write Access to Freed Host Kernel Memory
CVE-2026-89775 exposes freed host memory to the guest virtual machine. When the host kernel frees memory pages, the KVM code on ARM64 systems fails to properly isolate that memory from guest access. The guest VM can subsequently read the contents of freed host memory and write to those memory regions.
Freed memory often contains residual data from prior kernel operations—credentials, cryptographic keys, process memory, or kernel data structures. A guest VM reading freed host memory can extract sensitive information that should be isolated by the hypervisor boundary.
Write access to freed host memory is even more dangerous. If the guest can write to memory that the host kernel later reallocates for a critical data structure or process, the guest can corrupt host memory and potentially achieve code execution on the host system.
Researcher States Flaw Enables Guest-to-Host Code Execution
The researcher who disclosed CVE-2026-89775 stated the vulnerability can be used to escape the guest environment and execute code on the host machine. This assessment elevates the flaw from an information disclosure issue to a full VM escape vulnerability.
VM escape vulnerabilities represent the highest severity class of virtualization flaws. The entire purpose of a hypervisor is to isolate guest VMs from the host and from each other. A guest that can execute code on the host breaks the fundamental security boundary, gaining access to all VMs running on that host, the host’s storage and network, and any credentials or data the host manages.
For cloud providers, VM escape vulnerabilities allow tenant-to-tenant attacks. A malicious customer could rent a VM, exploit CVE-2026-89775 to escape to the host, and then attack other tenants’ VMs running on the same physical server.
Linux Kernel Maintainers Notified, Patches Expected
Linux kernel maintainers have been notified of CVE-2026-89775. Patches are expected in upcoming kernel releases, though no specific version or timeline was provided in the September 22 disclosure.
The disclosure recommends that users disable nested virtualization on ARM64 KVM hosts if it is not required. Disabling nested virtualization eliminates the attack surface for CVE-2026-89775, though it also removes the capability for guests to run their own VMs.
Mitigation Requires Disabling Nested Virtualization Until Patch Available
Organizations running ARM64 KVM hosts with nested virtualization enabled should evaluate whether nested virtualization is operationally necessary. If guests do not require the ability to run their own VMs, disabling the feature mitigates CVE-2026-89775 without waiting for a kernel patch.
For environments where nested virtualization is required—cloud providers offering nested virtualization as a feature, development platforms supporting virtualization testing, or security research labs—disabling the capability is not an option. These organizations must apply kernel patches as soon as they become available and accept the exposure window until patches are deployed.
Implications for ARM64 Cloud and Virtualization Infrastructure
CVE-2026-89775’s impact centers on ARM64 cloud infrastructure and virtualization platforms. Cloud providers that offer ARM64 instances with nested virtualization enabled face tenant-to-host escape risk. A malicious tenant could exploit the vulnerability to compromise the hypervisor host and access other tenants’ VMs or data.
The specificity of the vulnerability—ARM64 processors, KVM hypervisor, nested virtualization enabled—narrows the affected population compared to vulnerabilities affecting all hypervisors or all architectures. However, ARM64 adoption in cloud infrastructure has grown significantly, particularly for cost-sensitive workloads and power-efficient computing.
Organizations operating ARM64 KVM environments should identify which hosts have nested virtualization enabled, assess whether disabling the feature is operationally feasible, and prioritize kernel updates when patches are released. Hosts running untrusted or multi-tenant workloads require the most urgent action, as the risk of deliberate exploitation is highest in those environments.
The researcher’s public disclosure without a patch available creates an exposure window. Organizations aware of the vulnerability but unable to disable nested virtualization must monitor for exploitation attempts, review guest VM activity for anomalous host memory access patterns, and prepare to apply patches immediately upon release.
