Re: [RFD PATCH 3/5] sched/cpufreq_schedutil: make worker kthread be SCHED_DEADLINE

[Juri Lelli]

On 30/03/17 22:22, Rafael J. Wysocki wrote:
> On Thursday, March 30, 2017 08:50:11 AM Vikram Mulukutla wrote:
> > 
> > > OK
> > > 
> > > So there are two pieces here.
> > > 
> > > One is that if we want *all* drivers to work with schedutil, we need to 
> > > keep
> > > the kthread for the ones that will never be reworked (because nobody 
> > > cares
> > > etc).  But then perhaps the kthread implementation may be left alone 
> > > (because
> > > nobody cares etc).
> > > 
> > > The second one is that there are drivers operating in-context that work 
> > > with
> > > schedutil already, so I don't see major obstacles to making more 
> > > drivers work
> > > that way.  That would be only a matter of reworking the drivers in 
> > > question.
> > > 
> > > Thanks,
> > > Rafael
> > 
> > There are some MSM platforms that do need a kthread and would love to 
> > use
> > schedutil. This is all mainly due to the point that Vincent raised; 
> > having
> > to actually wait for voltage transitions before clock switches. I can't
> > speak about the future, but that's the situation right now. Leaving the
> > kthread alone for now would be appreciated!
> 
> I was not arguing for removing the kthread (quite opposite rather).
> 
> My point was that *if* it is viable to rework drivers to operate in-context,
> that would be the way to go IMO instead of messing up with the kthread thing.
> 

Right, I agree. Problem is that in principle we might still want to use
DEADLINE with the other platforms (MSM being a perfect example), so IMHO
we should still try to find a solution for the kthread anyway.