CVE-2026-68114: Linux Kernel DOS Vulnerability

CVE-2026-68114 Overview

CVE-2026-68114 affects the Linux kernel's AMD GPU driver, specifically the drm/amdgpu/gfx12.1 component. The vulnerability stems from the use of BUG_ON() macros in code paths that do not warrant a full kernel panic. Maintainers replaced these calls with WARN_ON() to preserve system stability when the underlying condition is triggered. The fix was cherry-picked from upstream commit e4d99e04b2e9b13b97d3b17804c735f62689db23 and applied to stable trees.

Critical Impact
Triggering the affected code path caused an immediate kernel panic through BUG_ON(), resulting in a denial-of-service condition on systems using AMD GFX 12.1 graphics hardware.

Affected Products

Linux kernel builds including the drm/amdgpu driver with GFX 12.1 support
Stable kernel branches prior to the referenced fix commits
Systems using AMD graphics hardware supported by the gfx12.1 code path

Discovery Timeline

2026-08-10 - CVE-2026-68114 published to NVD
2026-08-10 - Last updated in NVD database

Technical Details for CVE-2026-68114

Vulnerability Analysis

The vulnerability resides in the drm/amdgpu/gfx12.1 driver code within the Linux kernel graphics stack. The affected code used the BUG_ON() macro to assert runtime conditions. When BUG_ON() evaluates to true, the kernel executes a BUG() call, which triggers an oops and typically renders the system unusable until reboot.

The maintainer commit message states plainly that there is no need to crash the kernel for the checked conditions. The affected assertions guarded recoverable states, not conditions that would corrupt memory or compromise integrity. Replacing BUG_ON() with WARN_ON() allows the kernel to log a stack trace and continue running.

This class of issue is categorized as a kernel-level denial of service. An unprivileged local trigger of the guarded condition would take down the running kernel, affecting availability for all users on the system.

Root Cause

The root cause is defensive programming that escalated recoverable states to fatal assertions. BUG_ON() should be reserved for conditions where continued execution would cause memory corruption or security violations. In gfx12.1, the guarded conditions did not meet that threshold, yet BUG_ON() was used, producing a kernel panic on trigger [CWE-617: Reachable Assertion].

Attack Vector

The attack vector depends on whether the guarded code path can be reached through unprivileged user actions such as ioctl calls against the DRM device node, GPU submission queues, or workloads issued to the affected AMD hardware. Reaching the assertion produces a kernel panic and full system loss of availability. See the upstream fix commit for the exact code paths affected.

Detection Methods for CVE-2026-68114

Indicators of Compromise

Kernel panic messages referencing amdgpu and gfx_v12 symbols in the crash trace
Unexpected system reboots on hosts running AMD GFX 12.1 hardware with vulnerable kernels
dmesg output containing BUG_ON invocations from the drm/amdgpu subsystem prior to a hang

Detection Strategies

Inventory Linux hosts and compare running kernel versions against the fixed stable release that includes commits 1c27e889fa16 and 6560e6bd7612
Monitor system logs and crash dumps for repeated amdgpu oops signatures on GPU-equipped workloads
Correlate GPU workload submissions with subsequent kernel crashes on affected hardware

Monitoring Recommendations

Forward dmesg and /var/log/kern.log output to a centralized log platform for pattern analysis
Alert on any Kernel panic or BUG: events originating from graphics driver modules
Track uptime anomalies on GPU compute nodes to identify recurring driver-induced crashes

How to Mitigate CVE-2026-68114

Immediate Actions Required

Upgrade to a Linux kernel release that includes commits 1c27e889fa162bc3590de0942237d2ccec96b765 and 6560e6bd76127844e39f09fa591c2791dc7932e8
Apply the vendor-supplied kernel package updates on all systems running AMD GFX 12.1 hardware
Reboot affected systems to activate the patched kernel image

Patch Information

The fix replaces BUG_ON() with WARN_ON() in the drm/amdgpu/gfx12.1 driver. Reference the Linux kernel stable commit and the cherry-picked upstream commit for details.

Workarounds

Restrict access to DRM device nodes under /dev/dri/ to trusted users where feasible
Avoid running untrusted GPU workloads on unpatched systems with AMD GFX 12.1 hardware
Where the GPU is not required, blacklist the amdgpu module until patched kernels are deployed

bash

# Verify kernel version and check for the fix
uname -r

# On distributions using apt
sudo apt update && sudo apt upgrade linux-image-generic

# On distributions using dnf
sudo dnf update kernel

# Reboot to load the patched kernel
sudo systemctl reboot