On Thu, 2017-07-13 at 18:23 +0800, Ming Lei wrote: > Please discuss the congestion control in patch 4 and 5. Hello Ming, Since there is a significant risk that this patch series will introduce performance and/or functional regressions, I will wait with reviewing this patch series further until it has been shown that there is no less risky alternative to realize the same performance improvement. I think Jens had already suggested a possible alternative, namely by only starting a queue if it really has to be started. Bart.