Technical impact
When a power supply device is unregistered while its callback code is still executing, the kernel frees the driver’s data before the callback completes, causing a use‑after‑free that can dereference a null or invalid pointer and crash the kernel. The crash manifests as an oops and can bring the entire system down, as demonstrated on an ARM platform during an I²C bus fault that triggered a power‑supply‑changed workqueue.
Affected Systems
All Linux kernel versions that have not incorporated the commit adding a read‑write semaphore between external_power_changed() and power_supply_unregister() are vulnerable. The flaw affects every architecture and board that uses the power‑supply framework and relies on drivers that may unregister during an ongoing callback; the affected vendor and product are Linux:Linux, covering all Linux kernel distributions.
Risk and Exploitability
The typical surface is local privileged code that can trigger a power supply removal or a buggy driver that causes rapid deregistration. The EPSS score is < 1% and the issue is not listed in CISA KEV, but the CVSS score is 4.1. The severity is equivalent to a kernel panic, which is a high‑impact denial‑of‑service. The likely attack vector is the execution of untrusted or incomplete driver code (or a device‑fault condition) that initiates unregistering; this inference comes from the description of the driver removal race condition. Administrators should treat this as a high‑risk local vulnerability.