CVE-2026-47491

A flaw was found in nvidia-driver. A local unprivileged attacker can trigger the improper release of memory resources in the kernel mode layer, leaving a memory mapping accessible after the underlying memory is reused. A successful exploit could allow an attacker to achieve privilege escalation, execute arbitrary code, tamper with system memory, or cause a Denial of Service (DoS).

This vulnerability is classified as having an Important impact because it allows a local, unprivileged attacker to trigger improper memory management in the kernel, potentially resulting in full kernel-level code execution and privilege escalation. Red Hat Enterprise Linux systems are exposed only when the proprietary NVIDIA GPU driver is installed and actively in use with NVIDIA hardware. Systems using default open-source display drivers, such as nouveau, or systems without NVIDIA hardware are not affected.

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.

On systems where GPU hardware acceleration is not required, prevent the NVIDIA kernel driver module from loading:

1. Blacklist the kernel module by creating a modprobe configuration file:

echo "blacklist nvidia" > /etc/modprobe.d/blacklist-nvidia.conf

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

2. On systems where the driver must remain active, restrict access permissions on NVIDIA device nodes (such as /dev/nvidia* and /dev/nvidiactl) so that only trusted administrative accounts or designated service groups can access them.

Caveats: Preventing the module from loading disables GPU hardware acceleration, graphical environments relying on the proprietary driver, and CUDA workloads. A system restart is required for module blacklisting to take effect if the driver is currently loaded.