Re: [PATCH 2/3] spi: sh-msiof: Convert to spi core auto_runtime_pm framework

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Magnus,

On Wednesday 12 March 2014 10:23:48 Magnus Damm wrote:
> On Wed, Mar 12, 2014 at 1:32 AM, Laurent Pinchart wrote:
> > On Tuesday 11 March 2014 17:23:37 Geert Uytterhoeven wrote:
> >> Hi Laurent,
> >> 
> >> On Tue, Mar 11, 2014 at 4:55 PM, Laurent Pinchart wrote:
> >> > Does this require drivers/sh/pm_runtime.c to be compiled in ?
> >> 
> >> Let's check...
> >> 
> >> My koelsch-legacy kernel has drivers/sh/pm_runtime.c compiled in.
> >> My koelsch-reference kernel hasn't.
> >> 
> >> However, under the -reference kernel many MSTP clocks (incl. MSIOF)
> >> seem to be enabled all the time, while under -legacy they are enabled and
> >> disabled on demand.
> > 
> > Is PM_RUNTIME enabled in both cases ?
> > 
> > There's something fishy in there that we should try to fix without too
> > further much delay. Ben Dooks has pointed out the problem a couple of
> > months ago, but the discussion on the mailing list just died.
> 
> Yes, Runtime PM is not working as expected in the multiplatform case,
> that is true. I propose that we keep Runtime PM disabled in the
> Kconfig for R-Car Gen2 for now to keep things simple.

Isn't it ? I thought it was only broken with regard to clocks, but I might be 
missing something.

> From my side I see it as two separate solutions. The short term fix is
> to temporarily work around drivers that depend on Runtime PM for clock
> control, I propose enabling selected clocks statically using the
> function introduced by this series:
> 
> [PATCH 00/03] ARM: shmobile: Break out and extend clock workaround
> http://www.spinics.net/lists/arm-kernel/msg310278.html
> 
> The long term fix I'm not sure sure about, but I trust Geert to figure
> it out. =)
> 
> Regardless, rushing to fix this "correctly" seems as useful to me as
> dead line driven DT development....

-- 
Regards,

Laurent Pinchart