The fault resides in the handling of the PowerPC performance monitoring unit within the Linux kernel. A negative event index causes the cleanup path to skip the matching get_cpu_var() before executing put_cpu_var(), which in turn generates a preempt_enable() call without a preceding preempt_disable(). On CONFIG_PREEMPT=y kernels this mismatch underflows the preempt_count counter, eventually triggering the "scheduling while atomic" BUG. The resulting instability can lead to corrupted scheduling state, erratic interrupt handling or system crashes, effectively creating a denial‑of‑service vector for local users with sufficient privileges to load or trigger the faulty kernel module.

Affected systems are Linux kernel builds running on PowerPC architectures that include the fsl_emb_pmu performance monitoring unit implementation. Any kernel that incorporates the legacy code path in powerpc/perf before the commit that applies the fix is vulnerable; newer releases that backport the change are not affected.

The CVSS score of 5.5 indicates medium severity, reflecting the potential for a kernel panic or denial‑of‑service. The EPSS score remains below 1%, pointing to a low probability of exploitation under current conditions. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog, indicating no confirmed public exploits yet. The likely attack vector remains local privilege escalation or kernel module loading by a user with root or equivalent rights.