On Thu, May 17, 2018 at 11:05:30AM +0200, Peter Zijlstra wrote: > Hurm.. I think this is also wrong. See how dequeue_rt_stack() is also > called from the enqueue path. Also, nothing calls cpufreq_update_util() > on the throttle path now. > > And talking about throttle; I think we wanted to check rt_queued in that > case.
Bah, clearly you also want to update on unthrottle.. but I think there's more buggered wrt throtteling.