CVE-2026-47533

A flaw was found in nvidia-driver. A local attacker with administrative privileges can trigger an array indexing error within the kernel-mode driver layer. Successful exploitation could allow the attacker to execute arbitrary code, escalate system privileges, access sensitive data, or cause a denial of service (DoS).

This vulnerability is rated as Moderate severity for Red Hat Enterprise Linux because exploitation requires local access with administrative privileges (PR:H) to interact with the kernel-mode driver, significantly constraining the attack surface. The flaw involves improper boundary checking on array indices within the proprietary NVIDIA display driver, potentially enabling an already privileged local adversary to execute arbitrary code in kernel space, corrupt memory, or cause a system crash. Systems that do not have NVIDIA GPU hardware or do not load the proprietary driver modules remain unaffected.

The CVSS score(s) listed for this vulnerability do not reflect the associated product's status, and are included for informational purposes to better understand the severity of this vulnerability.

If NVIDIA GPU acceleration is not required on the host, prevent the kernel module from loading by configuring modprobe:

1. Create a configuration file to blacklist and disable autoloading of the module:

echo "blacklist nvidia" > /etc/modprobe.d/disable-nvidia.conf
echo "install nvidia /bin/false" >> /etc/modprobe.d/disable-nvidia.conf

2. Attempt to unload the module if currently active:

modprobe -r nvidia

Caveats and Warnings:

• Disabling the NVIDIA kernel module prevents the proprietary driver from functioning, which disables GPU display rendering, 3D acceleration, and compute frameworks (such as CUDA).
• If the module is actively in use by graphical sessions or running processes, unloading it may fail and a full system reboot will be required for the configuration to take effect. Rebooting or unloading the module will terminate active GPU workloads and graphical sessions.