Linux Kernel RCU Tasks Stall Vulnerability in Threaded Busy Polling

A vulnerability in the Linux kernel's handling of threaded busypoll can lead to RCU (Read-Copy-Update) task stalls. This issue was observed while debugging a network interface card (NIC) driver, where enabling threaded busypoll caused bpftrace to hang during startup. The kernel's message log indicated that the RCU tasks grace period was significantly overdue, suggesting a stall in task processing. The root cause lies in the way threaded busypoll is managed, as it disrupts the normal iteration of the polling loop, preventing timely updates to the RCU grace period. As a result, tasks can become stalled, leading to potential performance issues or hangs in applications that rely on timely task processing.

Impact

The vulnerability causes RCU task stalls, which can lead to hangs in applications that depend on the RCU mechanism for task synchronization. This was notably observed with bpftrace, which became unresponsive when threaded busypoll was enabled.

Reproduction

To reproduce this issue, enable threaded busypoll on a network interface and observe the behavior of bpftrace during startup. The application will hang, and the kernel's message log will show signs of RCU task stalls, indicating that the busypoll is not being processed correctly.

Remediation

Users can apply the patch available in the Linux kernel stable tree to address this vulnerability. The patch modifies the handling of the RCU grace period in threaded busypoll, ensuring that it is updated correctly and preventing stalls.