Hello, On Tue, Jul 11, 2017 at 07:51:07PM +0900, Tetsuo Handa wrote: > I tried below change. It indeed reduced delays, but even with WQ_HIGHPRI, up to a > few seconds of delay is unavoidable? I wished it is processed within a few jiffies. The rescuer doesn't take long to kick in but there's only one in the entire system. If you have 64 cpus, that one rescuer is gonna be visiting each CPU servicing them one-by-one and the scheduling latencies and execution times accumulate. Again, this is a deadlock prevention mechanism. If these per-cpu tasks need to be executed on every CPU without latency regardless of memory pressure, it has to pre-allocate all the resources up-front - ie. use kthread_workers instead of workqueue. > By the way, I think it might be useful if delay of each work item is > printed together... Yeah, sure, please feel free to submit a patch. Thanks. -- tejun -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
