Linux Kernel RDMA/mlx5 Implicit ODP Deregistration Hang Vulnerability

A vulnerability in the Linux kernel's RDMA/mlx5 component can cause a hang during the deregistration of an implicit memory region (MR) parent. This issue arises because the reference count for the implicit MR parent is not properly managed, leading to a blockage when the parent is deregistered. The problem has been addressed by updating the code to correctly handle the reference count, preventing the hang during the deregistration process.

Impact

The vulnerability can lead to a denial-of-service condition, where a process is blocked for an extended period, causing potential disruptions in service or application performance.

Reproduction

The vulnerability can be reproduced by initiating a deregistration process for an implicit memory region parent in the RDMA/mlx5 component. If the reference count is incremented but not properly decremented, the process will become blocked, demonstrating the hang issue. This can be observed in the task manager, where the blocked task will show a state of 'D' (uninterruptible sleep) for more than 120 seconds.

Remediation

Users should update to the latest version of the Linux kernel where this vulnerability has been fixed.