The Linux kernel RT scheduler contains logic that pushes high‑priority real‑time tasks to other CPUs via inter‑processor interrupts (IPIs). When the kernel is compiled without the PREEMPT_RT patch, this push mechanism can incorrectly schedule tasks while a softirq is still executing, causing the CPU to repeatedly receive IPIs. The cycle prevents the softirq from returning to task context, locking the CPU in an IPI handling loop and producing latency spikes of up to 500 µs. The flaw is a race condition in RT task migration that results in a denial of service for workloads sensitive to CPU availability.

All Linux kernel builds that include the RT_PUSH_IPI feature, including mainstream stable releases and custom kernels with real‑time scheduling, are impacted when compiled without the PREEMPT_RT patch. The feature is enabled by default in these builds, so any mainstream kernel that does not use the PREEMPT_RT patch will be vulnerable unless the RT_PUSH_IPI configuration is disabled. Kernels that are compiled with PREEMPT_RT enabled are not affected because the IPI logic is handled differently.

The CVSS score of 7.5 and an EPSS score of less than 1 % indicate a moderate severity but low likelihood of exploitation. The vulnerability does not provide a remote trigger; it requires a specific environment in which heavy network traffic generates frequent softirqs and the system runs real‑time tasks. The lack of listing in the CISA KEV catalog further supports a low threat level. Attackers would need privileged access or a kernel module to force the required load pattern, so the risk is primarily performance degradation under a narrow set of conditions rather than widespread exploitation.