Linux Flaw Lets VMs Read Neighbors' Memory
TL;DR: A critical flaw in Linux KVM lets a virtual machine escape and read the host server's memory. On cloud providers using AMD chips, this could expose sensitive data belonging to other customers on the same physical machine. Patches are available.
Key facts
- Category
- Cybersecurity
- Impact
- Critical
- Published
- Source
- The Hacker News
Full summary
A flaw in Linux KVM allows virtual machines on AMD-powered hosts to read the server's memory, potentially exposing other customers' sensitive data.
Security researchers have discovered a critical vulnerability in KVM, the virtualization technology built directly into the Linux kernel. According to The Hacker News, the flaw allowed them to read approximately 127 GB of a cloud server's main memory from within a standard virtual machine they controlled. The demonstration took place on the DigitalOcean cloud platform, highlighting a real-world risk to public cloud infrastructure. Tracked as CVE-2026-99312, the vulnerability specifically affects host servers running on AMD processors with a feature called nested virtualization enabled. The ability for one tenant's virtual machine to read memory outside its designated boundary is a fundamental breach of cloud security, as that memory could contain sensitive data from other customers sharing the same physical hardware.
This type of vulnerability is known as a hypervisor escape. The hypervisor, in this case KVM, acts as a digital landlord, creating and strictly isolating multiple virtual machines (tenants) on a single physical server. Its primary job is to ensure that no VM can access resources—especially memory—that don't belong to it. This flaw effectively breaks that isolation. It exploits a weakness in how KVM manages nested virtualization on AMD chips. Nested virtualization allows a VM to run its own hypervisor, a complex feature often used for development and testing environments. The bug appears to create a condition where the guest VM can trick the host hypervisor into giving it access to physical memory addresses outside of its sandboxed environment, allowing it to read raw data from the host and, by extension, other VMs.
The discovery underscores a persistent tension in cloud computing: the drive for more powerful features versus the need for uncompromising security. While nested virtualization enables flexible and complex architectures, it also increases the attack surface of the hypervisor. This incident joins a history of notable hypervisor escapes, such as VENOM and Cloudburst, which have periodically reminded the industry that the isolation layer is not infallible. The security of the entire multi-tenant cloud model, from the largest public clouds to private enterprise deployments, rests on the integrity of this virtualization layer. A flaw in a foundational component like KVM, which is used extensively across the industry, represents a systemic risk that can undermine the trust customers place in shared infrastructure.
Fortunately, the Linux kernel development community has already released patches to fix the vulnerability. For organizations managing their own infrastructure using KVM on AMD hardware, the immediate action is to apply these kernel updates as soon as possible. For customers of public cloud providers, the responsibility lies with the provider to patch their fleet of servers. Customers should seek confirmation from their vendors that the necessary mitigations are in place. This event will likely spur increased scrutiny of nested virtualization implementations across all processor architectures, as security researchers will now be looking for similar weaknesses. It serves as a critical reminder that robust security requires constant vigilance, from hardware-level features to the software that manages them, and that prompt patching remains one of the most effective defenses against emerging threats.
Why it matters
This vulnerability breaks the fundamental isolation boundary between a guest VM and its host, a core security promise of virtualization. For any organization running workloads on shared infrastructure with AMD processors, it represents a critical risk of data exposure to other tenants, undermining multi-tenant security models.
Business impact
A hypervisor escape of this nature creates a significant data breach risk for both cloud providers and their customers. It can lead to regulatory fines, loss of customer trust, and expensive incident response efforts. Companies must ensure their infrastructure providers have applied the necessary patches.
⚡ Action needed
Security patches are available and should be applied immediately.
Action checklist
- 1Identify all Linux hosts running KVM on AMD processors.
- 2Determine if nested virtualization is enabled on these hosts.
- 3Consult your Linux distribution's security advisories for the patched kernel version.
- 4Schedule and apply the kernel updates to all affected hosts.
- 5Reboot hosts to ensure the new kernel is active.
Tags
Related on Notifire
Related stories
Primary source: The Hacker News
