In the Linux kernel’s KASAN subsystem a race condition exists between cache shrink operations and CPU hotplug events. When a CPU is taken offline while a slab cache is being shrunk, objects are moved to that CPU’s shrink list but never returned to the slab allocator after the CPU goes offline. The kernel then reports that the cache cannot be destroyed cleanly, indicating that memory slabs remain allocated and leak over time. The flaw does not provide a direct code‑execution path but can lead to gradual memory exhaustion, kernel instability, and eventual service disruption.

The defect appears only in kernels that enable CONFIG_KASAN_GENERIC, which builds the generic KASAN runtime to detect memory corruption. The affected product is the Linux kernel itself; any build that includes KASAN with CPU hotplug support is subject to the issue. No specific release version is listed, but the commit history in the supplied URLs shows that the fix was merged into the mainline kernel and is available in current stable releases.

The race requires a CPU hotplug event to occur concurrently with cache shrink, a condition that is unlikely in normal operation and for which no public exploit exists. The EPSS score is not provided and the vulnerability is not catalogued by CISA KEV, suggesting no known active exploitation. Nevertheless, the potential for memory exhaustion is high, and the impact classification indicates a severe vulnerability if left unpatched.