Skip to main content
CybersecurityVulnerability Management

Google Cloud Warns of AMD Turin CPU Flaws That Leak Data Across VM Processes

Rows of server racks and computer equipment in a data center, with a single server node and abstract CPU representation in…

"Rebooting the guest OS isn't enough," Google Cloud warns — a short sentence that drives the operational action required to protect virtual machines running on AMD's Turin (Zen 5) processors.

Which servers and CVEs are affected

Google Cloud identified three Transient Scheduler Attack (TSA) vulnerabilities that can leak data across processes inside a virtual machine on Turin-based instances. Two variants, TSA-SQ (CVE-2024-36350) and TSA-L1 (CVE-2024-36357), were previously disclosed on July 8, 2025, for earlier AMD EPYC processors; a new variant, CVE-2026-0004, affects the processor's L1 data cache and is specific to Zen 5 (Turin).

The affected Google Cloud machine series are C4D, N4D, H4D, Z4D, and G4, including bare-metal instances of C4D and Z4D. Turin is identified in the bulletin as the code name for fifth-generation AMD EPYC processors.

Google Cloud's guidance for VMs, bare metal, GKE and Cloud Run

Google Cloud's security bulletin lays out distinct actions by service:

  • VMs: Install the guest OS vendor's fix when it is released. Starting the week of this bulletin, Google Cloud will update the CPU information that VMs report so a patched guest kernel can enable mitigations; after that update reaches a VM, customers must stop and start the VM — a simple OS reboot will not suffice.
  • Bare metal instances: Operators running their own OS or hypervisor on C4D or Z4D bare metal instances should apply their vendor's fixes when available.
  • Hypervisors inside a VM: The same CPU information update will allow patched guest kernels to turn on mitigations; after the update arrives, the VM must be stopped and started.
  • GKE: Updated node images for Turin-based bare metal instances will be made available, and upgrading nodes will pick up the CPU information update; as of today the bulletin does not list those updated images as available.
  • Cloud Run: No action is required.

Google, AMD, Xen Project, CISA: classifications and real-world evidence

Google rates the issue Medium in its bulletin. AMD classifies all three flaws as local attacks with high attack complexity that require only low privileges, according to its Zen 5 bulletin and the CVE-2026-0004 advisory.

As of today, none of the three CVEs appears in the KEV catalog of exploited vulnerabilities maintained by the U.S. Cybersecurity and Infrastructure Security Agency (CISA), and neither Google Cloud nor AMD reports any exploitation in their public bulletins.

The Xen Project, which maintains the Xen hypervisor, has named the new variant TSA-Fill and issued hypervisor patches in its advisory. The Xen Project also corrected an earlier statement, noting: "AMD had previously said that Zen5 CPUs were not vulnerable to TSA-SQ and TSA-L1."

Kernel and distribution status: Linux source, Debian, linux-firmware

A kernel commit that mitigates TSA on Zen 5, including the new variant, is present in the Linux kernel source tree. Debian has applied a fix for CVE-2026-0004 in its Linux kernel only in the unstable branch; Debian's security tracker continues to list the kernels in bookworm, trixie, and forky as vulnerable.

Xen Project notes that microcode addressing all TSA flaws on Zen 5 has been included in linux-firmware since September 23. Administrators and maintainers will therefore need combinations of microcode, platform firmware, hypervisor patches, and kernel updates to reach full mitigation.

AMD's technical guidance and operational trade-offs

AMD states that for Turin servers on customer-owned hardware both platform firmware and OS updates are required to mitigate the flaws fully. The company warns that mitigations may affect performance and advises administrators to weigh the risk to their particular workloads when applying fixes, according to AMD's technical guidance.

What this means for technologists, cloud customers, and administrators

  • Technologists and security teams: Expect a multi-component remediation — microcode/platform firmware, kernel patches, and hypervisor updates — and test performance impacts before wide deployment.
  • Cloud customers and procurement leaders: For Google Cloud VMs, monitor your OS vendor for the guest kernel fix, then watch for Google’s CPU information update and perform a stop-and-start — not just a reboot — to enable mitigations on affected instances.
  • Administrators of on-prem Turin servers: Plan to coordinate platform firmware and OS updates and evaluate performance trade-offs against your risk profile; Xen hypervisor operators should apply the Xen advisory patches and confirm linux-firmware placement of microcode.

Two practical operational gaps remain in the record provided: Google Cloud lists no interim actions for VMs whose OS vendor has not yet shipped a fix, and the bulletin does not specify how customers can confirm when the CPU information update has actually reached a VM. Those are the next concrete points affected teams will need resolved as they move from notice to remediation.

Source: https://thehackernews.com/2026/11/google-cloud-warns-three-amd-turin-cpu.html