On 9 Jan 2023 13:33:16 +0000 Valentin Schneider <vschneid@xxxxxxxxxx> > It has been reported that isolated CPUs can suffer from interference due to > per-CPU kworkers waking up just to die. > > A surge of workqueue activity during initial setup of a latency-sensitive > application (refresh_vm_stats() being one of the culprits) can cause extra > per-CPU kworkers to be spawned. Then, said latency-sensitive task can be > running merrily on an isolated CPU only to be interrupted sometime later by > a kworker marked for death (cf. IDLE_WORKER_TIMEOUT, 5 minutes after last > kworker activity). Given no scheduling in userspace, how is the latency-sensitive task interrupted by woken kworker? If it is due to wakeup preempt then disabling it on isolated CPU for user mode makes more sense, but that should add change to scheduler instead.
