When looking up the pool_workqueue to use for an unbound workqueue, workqueue assumes that the target CPU is always bound to a valid NUMA node. Howev
↧