A vulnerability in the Linux kernel's io_uring subsystem allows a race condition where a foreign task can enqueue I/O work to a freed or null I/O work queue. This occurs because an IOPOLL ring without DEFER_TASKRUN can be reaped by a different task than the one that submitted the I/O, leading to potential use-after-free scenarios. The issue has been addressed by deferring foreign userspace enqueues to the owning task using a dedicated task_work callback.