CVE-2022-34675

What it actually does

The Virtual GPU Manager in the NVIDIA Display Driver for Linux fails to check the return value following a null-pointer dereference, allowing the dereference to proceed unchecked and resulting in a crash or denial of service.

Why it matters to the business

Virtualized GPU workloads on Linux hosts using Citrix Hypervisor, Red Hat KVM, or NVIDIA cloud gaming can experience abrupt unavailability, causing downtime for graphics or compute VMs without data loss but with operational disruption.

How an attack unfolds

An attacker with local access on an affected Linux system supplies crafted input through GPU driver interfaces to reach the unchecked return-value path inside the vGPU Manager, triggering the null dereference and crashing the driver.

How to fix it

Apply the latest NVIDIA Display Driver updates for all listed affected products (GeForce, Quadro, RTX, Tesla, etc.) as soon as vendor patches become available; until then restrict local access to GPU management interfaces and monitor for driver instability.

How to confirm the fix worked

Verify installed driver versions against NVIDIA security bulletins, review release notes for the null-pointer handling fix, and perform functional testing of vGPU workloads to confirm continued operation under load.

Typical attack path

1. Obtain local user or process access on the Linux host
2. Interact with exposed GPU driver or vGPU management interfaces
3. Submit input that exercises the null-pointer code path
4. Trigger the unchecked return value leading to driver crash

How to detect it

Inspect system and kernel logs for NVIDIA driver crash signatures or vGPU Manager faults around the time of service loss; correlate with local process activity that accesses GPU device nodes.

What your security team should do

• Treat as priority patch for any Linux hosts running vGPU or NVIDIA display drivers
• Limit local privileges on systems exposing GPU management interfaces
• Add driver crash and vGPU availability alerts to existing monitoring
• Review and reduce the attack surface of virtualized GPU deployments