CVE-2026-47561

A flaw was found in the NVIDIA GPU display driver. A local, unprivileged user can exploit this vulnerability by submitting an input/output control (ioctl) request with an unvalidated buffer size to the driver's kernel layer. This causes an out-of-bounds write in kernel memory, which may allow the attacker to execute arbitrary code with elevated privileges, crash the system leading to a Denial of Service (DoS), or tamper with sensitive data.

This vulnerability is evaluated as having an Important severity rather than Critical because exploitation requires existing local access, but allows an unprivileged user to achieve complete kernel compromise and elevated privileges. Proprietary NVIDIA display drivers are not shipped or enabled by default in standard Red Hat Enterprise Linux environments, limiting exposure to systems where these specific driver packages have been explicitly deployed on NVIDIA hardware. Deployments utilizing default open-source graphics drivers or environments without NVIDIA accelerators are 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 GPU acceleration or proprietary display driver functionality is not strictly required, prevent the NVIDIA kernel module from loading:

1. Blacklist the module by creating a configuration file in /etc/modprobe.d/:

echo 'install nvidia /bin/true' > /etc/modprobe.d/disable-nvidia.conf

echo 'blacklist nvidia' >> /etc/modprobe.d/disable-nvidia.conf

2. Regenerate the initial ramdisk and restart the host:

dracut --force

systemctl reboot

Warning: A system reboot is required to completely unload the active module from memory.

Caveats: Disabling this driver module prevents hardware-accelerated graphics and compute workloads from accessing NVIDIA GPU devices. Applications relying on CUDA or proprietary display features will fail to function.