Executive brief
A vulnerability in the Linux kernel's Mellanox network driver can cause the system to hang indefinitely during certain hardware reset or shutdown operations. This occurs when the network card is configured in a high-availability 'LAG' mode, where a failure in one part of the hardware prevents the system from properly cleaning up resources. This can lead to a denial-of-service where the server becomes unresponsive during maintenance or hardware failure events.
Technical details
A race condition exists in the mlx5_ib driver when operating in Link Aggregation (LAG) mode during a firmware reset. In this configuration, the bond device is registered only on the master, causing it to miss 'sys_error' events from the slave device. During a device unload or firmware reset, the driver may wait indefinitely for User Mode Registration (UMR) completions that will never arrive because the slave device is unresponsive while the master has not yet entered an error state. This results in a kernel hang during the teardown process. The fix involves registering a 'sys_error' notifier earlier in the device lifecycle to ensure error events are properly propagated during teardown.
Affected products
- Linux Linux Kernel 6.13, 6.12, 6.6, 6.1
Timeline
- 2026-01-13: patched: Initial fix authored by Nvidia engineers.
- 2026-05-27: advisory: CVE-2026-45973 published.