Linux Kernel Task Affinity and Cpuset Update Race Condition Vulnerability

A vulnerability in the Linux kernel's scheduling component has been addressed. The issue arose in the 'sched_setaffinity' function, where a warning was introduced to signal a race condition between per-task affinity assignments and cpuset updates. This race could cause a task's affinity to fall outside the bounds of its cpuset, which is normally acceptable since the system can revert to using the cpuset mask. However, the warning was triggered if the cpuset mask completely lacked overlap with the requested task affinity, a condition that can be easily manipulated. The warning was reproduced by placing a process ID within a cpuset cgroup and having one thread continuously alter the cpuset's CPU allocation while another thread adjusted the process's affinity, creating a conflict that triggered the warning.

Impact

The vulnerability could lead to unnecessary warnings being generated, indicating a misalignment between task affinity and cpuset allocations, which could disrupt scheduling decisions.

Reproduction

The warning can be reproduced by assigning a process ID to a cpuset cgroup and then using one thread to repeatedly change the cpuset's CPU assignments, while another thread simultaneously modifies the process's affinity. This creates a race condition that triggers the warning by causing the affinity to conflict with the cpuset's CPU allocations.