CVE-2024-35959

What it actually does

Missing acquisition of state_lock in mlx5e_selq_cleanup leads to unprotected RCU dereference in mlx5e_selq_apply during mlx5e_priv_init error paths, triggering lockdep warnings and potential races in Mellanox Ethernet driver initialization.

Why it matters to the business

Kernel warnings or instability can occur during boot or network device setup on systems using affected Mellanox hardware with RDMA/InfiniBand over Ethernet, potentially delaying service availability in high-performance networking environments.

How an attack unfolds

An attacker triggers mlx5e_priv_init failure during module load or client registration (such as ipoib_add_one) to reach the flawed cleanup path, exercising the unprotected RCU access without holding the required lock.

How to fix it

Update the Linux kernel to a version containing commit ad26f26abd353113dea4e8d5ebadccdab9b61e76 or later; rebuild and deploy the fixed kernel across affected systems.

How to confirm the fix worked

Verify the presence of commit ad26f26abd353113dea4e8d5ebadccdab9b61e76 in the kernel source or check that the warning no longer appears in dmesg during mlx5e_priv_init failure scenarios.

Typical attack path

1. Load or register affected mlx5e network driver under conditions that cause mlx5e_priv_init to fail
2. Trigger mlx5e_selq_cleanup which calls mlx5e_selq_apply without state_lock held
3. Observe or exploit resulting suspicious RCU usage and potential race condition

How to detect it

Search kernel logs (dmesg, /var/log/kern.log) for 'WARNING: suspicious RCU usage' combined with 'mlx5e_selq_apply' and stack traces referencing mlx5e_priv_init or ipoib_intf_init; correlate with driver load events.

• WARNING: suspicious RCU usage
• mlx5e_selq_apply+0x94/0xa0
• mlx5e_selq_cleanup+0x3a/0x60
• drivers/net/ethernet/mellanox/mlx5/core/en/selq.c:124

What your security team should do

• Apply the kernel fix containing ad26f26abd353113dea4e8d5ebadccdab9b61e76 as soon as possible given the 877 days since publication
• Monitor kernel logs for 'suspicious RCU usage' messages involving mlx5e_selq_apply during driver initialization
• Test network and RDMA functionality after patching to confirm stability