CVE-2024-0108: Nvidia Jetson Linux DOS Vulnerability

CVE-2024-0108 Overview

CVE-2024-0108 is a high-severity vulnerability in NVIDIA Jetson Linux affecting the NvGPU component. The flaw resides in the GPU Memory Management Unit (MMU) mapping code, where error handling paths fail to clean up after a failed mapping attempt [CWE-755]. A local authenticated attacker can trigger the faulty error path to corrupt GPU memory state. Successful exploitation can lead to denial of service, arbitrary code execution, and privilege escalation across security boundaries. The vulnerability affects the entire Jetson product family, including AGX Xavier, Xavier NX, TX1, TX2, and Nano platforms used in robotics, automotive, and edge AI deployments.

Critical Impact
Local attackers with low privileges can escalate privileges, execute arbitrary code, or crash GPU-dependent workloads on affected Jetson edge devices.

Affected Products

NVIDIA Jetson Linux (all versions prior to the fix)
NVIDIA Jetson AGX Xavier series (8GB, 16GB, 32GB, 64GB, Industrial)
NVIDIA Jetson Nano, Nano 2GB, TX1, TX2, TX2 NX, TX2i, Xavier NX (including 16GB)

Discovery Timeline

2024-08-08 - CVE-2024-0108 published to the National Vulnerability Database
2024-09-16 - Last updated in NVD database

Technical Details for CVE-2024-0108

Vulnerability Analysis

The vulnerability lives in the NvGPU driver's GPU MMU mapping code path on NVIDIA Jetson Linux. When the kernel attempts to map memory into the GPU's virtual address space and that mapping operation fails, the error handler does not properly unwind the partial state it created. This improper handling of exceptional conditions is classified as [CWE-755].

The consequence is a corrupted GPU MMU state, which may include stale page table entries, dangling references to freed pages, or inconsistent reference counts. An attacker with local access and the ability to invoke GPU ioctls can repeatedly trigger the failed mapping path to manipulate kernel memory state.

Because the GPU driver runs at high privilege and interacts with DMA-capable hardware, residual state from the failed cleanup can be weaponized for memory corruption. The scope change (S:C) in the CVSS vector reflects that exploitation can affect resources beyond the originally authorized component, enabling cross-boundary privilege escalation.

Root Cause

The root cause is incomplete error path handling in the NvGPU MMU mapping logic. The mapping function allocates and modifies kernel and GPU page table structures, but the failure cleanup routine omits steps required to revert these modifications. This leaves the system in an inconsistent state that violates kernel memory safety invariants.

Attack Vector

Exploitation requires local access with low privileges on the Jetson device. An attacker invokes GPU driver interfaces, typically through /dev/nvhost-* device nodes accessible to graphics-capable user accounts, and supplies inputs that force the MMU mapping routine to fail. By repeating the failure condition, the attacker steers GPU and kernel memory into a corrupted state suitable for code execution or privilege escalation. No user interaction is required, and the attack does not require a network path.

No public proof-of-concept exploit code is available. The vulnerability mechanism is described in prose only. Refer to the NVIDIA Support Article for vendor-supplied technical details.

Detection Methods for CVE-2024-0108

Indicators of Compromise

Unexpected kernel log entries from the nvgpu driver referencing MMU mapping failures, page table errors, or fault recovery routines
GPU process crashes, hangs, or nvhost watchdog resets followed by privileged process anomalies
Unauthorized local users invoking GPU device nodes (/dev/nvhost-gpu, /dev/nvhost-as-gpu) outside expected workload patterns

Detection Strategies

Monitor dmesg and journalctl output for repeated nvgpu MMU fault messages or mapping error returns, which can indicate exploitation attempts
Audit process activity around GPU device files to identify unexpected processes opening nvhost interfaces
Establish a baseline of legitimate GPU workloads on Jetson devices and alert on deviations such as new binaries issuing GPU ioctls

Monitoring Recommendations

Forward Jetson kernel logs to a centralized logging platform for correlation across the edge device fleet
Track package versions of nvidia-l4t-kernel and nvidia-l4t-nvgpu to identify unpatched devices
Alert on sudden privilege escalations or new root-owned processes spawned shortly after GPU error log entries

How to Mitigate CVE-2024-0108

Immediate Actions Required

Apply the NVIDIA Jetson Linux update referenced in the NVIDIA Support Article to all affected Jetson devices
Inventory deployed Jetson hardware and confirm each device receives the patched BSP (Board Support Package)
Restrict shell and local login access on Jetson devices to trusted administrators and service accounts only

Patch Information

NVIDIA has released a fix for CVE-2024-0108 as part of a Jetson Linux security update. Affected operators should consult NVIDIA's advisory (Answer ID 5555) and update the Jetson Linux Board Support Package on all affected hardware models, including AGX Xavier, Xavier NX, TX1, TX2, Nano, and their variants. The patch corrects the error handling path in the NvGPU MMU mapping code so that failed mappings are fully unwound.

Workarounds

Limit membership in the video and render groups so only required service accounts can access GPU device nodes
Apply mandatory access control profiles (AppArmor or SELinux) to confine processes that interact with /dev/nvhost-* interfaces
Disable or remove untrusted local user accounts and disable interactive SSH for non-administrative users on production Jetson deployments

bash

# Restrict GPU device access to a dedicated group on Jetson Linux
sudo groupadd gpu-trusted
sudo chgrp gpu-trusted /dev/nvhost-gpu /dev/nvhost-as-gpu
sudo chmod 0660 /dev/nvhost-gpu /dev/nvhost-as-gpu

# Verify the installed Jetson Linux package version after patching
dpkg -l | grep nvidia-l4t-core