The Linux kernel scheduler can enter a costly retry loop when it attempts to migrate a task that has been marked migrate_disable(). The scheduler still keeps that task on the CPU’s pushable list and marks the runqueue as overloaded, so the RT balancer keeps trying to push it. Because the task cannot be moved, each push attempt fails and the loop repeats for tens of milliseconds, consuming up to half of the CPU’s time on the affected core. The loop never terminates, causing other tasks—including the currently running one—to receive far less CPU time and potentially never run. This behavior effectively denies service to real‑time workloads and other processes on that core.

The vulnerability is present in all Linux kernel versions that include the buggy scheduling logic before the patch referenced in the advisory. The affected vendor is Linux, product Linux kernel. No specific version ranges are listed in the input; the defect exists until the kernel is updated to incorporate the commit that fixes the push‑candidate selection loop.

The CVSS score is 5.5 and the EPSS score is <1%, and the vulnerability is not listed in the CISA KEV catalog, but the description shows that the bug can cause drastic CPU starvation and high system load. The exploit requires a task that is scheduled as a real‑time process with migrate_disable() enabled; while it may be difficult for an unauthenticated attacker to enforce this, privileged or locally compromised processes can trigger it. This is inferred from the description that the exploit depends on a migrate_disabled RT task. Because the loop is purely internal to the kernel scheduler, the attack vector is local and requires control over RT scheduling or the ability to spawn such tasks. This inference follows from the internal nature of the scheduler. The high CPU consumption and resultant denial of service make the risk significant, especially on systems that run critical real‑time workloads.