From mboxrd@z Thu Jan  1 00:00:00 1970
From: lorenzo.pieralisi@arm.com (Lorenzo Pieralisi)
Date: Tue, 7 Jul 2015 12:03:04 +0100
Subject: [RFC] arm: psci: implement cpuidle_ops
In-Reply-To: <20150702153253.GB14244@leverpostej>
References: <20150702111031.05f04184@xhacker> <5594F12F.6050401@linaro.org>
 <20150702161820.15dc9770@xhacker>
 <20150702153253.GB14244@leverpostej>
Message-ID: <20150707110304.GD4379@red-moon>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Thu, Jul 02, 2015 at 04:32:53PM +0100, Mark Rutland wrote:
> On Thu, Jul 02, 2015 at 09:18:20AM +0100, Jisheng Zhang wrote:
> > Dear Daniel,
> > 
> > On Thu, 2 Jul 2015 10:07:11 +0200
> > Daniel Lezcano <daniel.lezcano@linaro.org> wrote:
> > 
> > > On 07/02/2015 05:10 AM, Jisheng Zhang wrote:
> > > > Hi all,
> > > >
> > > > we'd like to use cpuidle-arm.c for both arm and arm64 with psci as backend. For
> > > > arm64, it works. But for arm, we miss cpuidle_ops. I want to add cpuidle_ops for
> > > > arm psci, I dunno whether this is the correct direction, could you please give
> > > > suggestions?
> > > 
> > > You should look at the macro
> > > 
> > > arch/arm/include/asm/cpuidle.h:
> > > 
> > > #define CPUIDLE_METHOD_OF_DECLARE(name, _method, _ops)
> > 
> > Yep, this is what I want to use. I want to use this MACRO to implement cpuidle_ops
> > for arm, but I dunno whether this is the correct direction. And I'd like to
> > reuse cpu_psci_cpu_init_idle() function in arch/arm64/kernel/psci.c to parse
> > the DT, but I'm not sure where to put cpu_psci_cpu_init_idle() so that can
> > be shared between arm and arm64. Also I noticed that there's a psci unification
> > work from Mark Rutland, does it make sense to put it in drivers/firmware/psci.c?
> 
> I think cpu_psci_cpu_init_idle() should live in drivers/firmware/psci.c.
> I also think that Lorenzo had an idea as to what should be done
> regarding unifying the arm and arm64 approaches, but he's currently away
> and I'm not all that familiar with cpuidle.

Yes, PSCI idle states initialization must be common between arm and arm64,
and also cpu_psci_cpu_suspend now that the cpu_suspend() API was made
common.

The drivers/firmware PSCI implementation for those hooks can be used to
inizialize the respective arm/arm64 structures then, as highlighted
in this thread. I will review the series.

Thanks,
Lorenzo