Re: [PATCH 15/18] drivers: firmware: psci: Support CPU hotplug for the hierarchical model
From: Sudeep Holla
Date: Mon Jun 10 2019 - 07:07:19 EST
On Mon, Jun 10, 2019 at 12:21:47PM +0200, Ulf Hansson wrote:
> On Fri, 7 Jun 2019 at 17:31, Sudeep Holla <sudeep.holla@xxxxxxx> wrote:
> > On Mon, May 13, 2019 at 09:22:57PM +0200, Ulf Hansson wrote:
> > > When the hierarchical CPU topology is used and when a CPU has been put
> > > offline (hotplug), that same CPU prevents its PM domain and thus also
> > > potential master PM domains, from being powered off. This is because genpd
> > > observes the CPU's attached device as being active from a runtime PM point
> > > of view.
> > >
> > > To deal with this, let's decrease the runtime PM usage count by calling
> > > pm_runtime_put_sync_suspend() of the attached struct device when putting
> > > the CPU offline. Consequentially, we must then increase the runtime PM
> > > usage count, while putting the CPU online again.
> > >
> > Why is this firmware/driver specific ? Why can't this be dealt in core
> > pm-domain ? I am concerned that if any other architectures or firmware
> > method decides to use this feature, this need to be duplicated there.
> What is the core pm-domain? Do you refer to the generic PM domain (genpd), no?
Sorry for my bad choice of names. I just wrote names as I understand
rather than looking for exact match. But yes, I meant generic place
where such ref-counting is done currently for other things.
> In such case, this is not the job of genpd, but rather the opposite
> (to *monitor* the reference count).
OK, I need to understand that then.
> > The way I see this is pure reference counting and is hardware/firmware/
> > driver agnostic and can be made generic.
> As stated in the another reply, I would rather start with having more
> things driver specific rather than generic. Later on we can always
> consider to move/split things, when there are more users.
> In this particular case, the runtime PM reference counting is done on
> the struct device*, that genpd returned via
> dev_pm_domain_attach_by_name(). And because
> dev_pm_domain_attach_by_name() is called from PSCI code, I decided to
> keep this struct device* internal to PSCI.
Sure, I understand your intent. I have just mentioned my thoughts/comments
as I reviewed.