CVE-2024-39468

What it actually does

Race condition in Linux kernel SMB client allows deadlock when cifs_tcp_ses_lock is held while calling cifs_put_smb_ses, causing the process to hang indefinitely during SMB session teardown.

Why it matters to the business

Systems using affected NVIDIA CUDA toolkit packages on Debian may experience hangs during networked file operations, leading to reduced availability for GPU-accelerated workloads and potential operational disruption until reboot.

How an attack unfolds

A local attacker triggers concurrent SMB session operations that contend on the lock ordering, causing the deadlock; no remote or unauthenticated path exists and no known threat actors have exploited it.

How to fix it

Upgrade nvidia-cuda-toolkit on Debian 13 and 14 to version 12.4.1-1 or later; rebuild or reinstall any dependent CUDA workloads after the package update.

How to confirm the fix worked

Confirm package version is 12.4.1-1 or newer via apt, then exercise SMB session creation and teardown under load to verify no deadlock occurs.

Typical attack path

1. Obtain local user access on an affected Debian system running nvidia-cuda-toolkit
2. Initiate multiple concurrent SMB/CIFS connections that exercise smb2_find_smb_tcon
3. Force session teardown paths while the lock is held to induce deadlock
4. Observe process or kernel thread hang affecting file I/O

How to detect it

Search kernel logs and dmesg for hung task warnings involving cifs_put_smb_ses or smb2_find_smb_tcon; inspect process lists for stuck CIFS worker threads.

What your security team should do

• Prioritize upgrade of nvidia-cuda-toolkit to 12.4.1-1 on all Debian 13/14 systems within standard change windows
• Monitor kernel logs for hung tasks related to cifs or smb2 after deployment
• Review any custom SMB mount scripts for potential lock-ordering assumptions