On 17-06-21, 14:22, Peter Zijlstra wrote: > On Thu, Jun 17, 2021 at 04:49:36PM +0530, Viresh Kumar wrote: > > Peter: Can you help here on this ? Lemme try to explain a bit here: > > > > We are starting an irq-work (in cppc cpufreq driver) from > > scheduler_tick()->arch_scale_freq_tick(). What will happen if the driver doesn't > > take care of CPU hotplug explicitly and make sure this work isn't queued again > > from the next tick. > > > > Is it important for user to make sure it gets rid of the irq-work during hotplug > > here ? > > irq-work is flushed on hotplug, see smpcfd_dying_cpu(). Thanks Peter. -- viresh