CVE-2024-26719

What it actually does

Race condition deadlock between fctx lock and irq lock during fence uevent processing in interrupt context within the nouveau driver.

Why it matters to the business

Systems using NVIDIA GPUs with the nouveau driver may experience kernel deadlocks causing system hangs or crashes, leading to operational downtime; high prevalence of Linux kernels amplifies exposure across server and desktop fleets.

How an attack unfolds

An attacker with local access triggers GPU fence operations that invoke uevent handling in the nouveau driver, forcing contention between the fctx and irq locks and resulting in a deadlock; no named threat actors or ransomware campaigns associated.

How to fix it

Update the Linux kernel to include commit cc0037fa592d56e4abb9c7d1c52c4d2dc25cd906 or later stable releases that contain the offload of fence uevents to a workqueue.

How to confirm the fix worked

Confirm presence of the fix commit in the kernel source tree or verify that running kernel packages are newer than the vulnerable versions listed for the affected git repository.

Typical attack path

1. Obtain local unprivileged access to a Linux system using nouveau
2. Perform operations that generate fence uevents in the nouveau driver
3. Induce simultaneous lock acquisition from interrupt and process contexts
4. Trigger deadlock causing kernel hang or crash

How to detect it

Inspect kernel logs and dmesg output for nouveau fence processing warnings, hung task messages involving fctx or irq locks, or workqueue-related stack traces indicating deadlock conditions.

What your security team should do

• Prioritize kernel updates on all systems using the nouveau driver
• Test updated kernels in staging environments before broad deployment
• Review boot logs and dmesg for nouveau-related hung task warnings as a temporary detection measure
• Evaluate switching to proprietary NVIDIA drivers where feasible to reduce reliance on nouveau