Clock & PM breakage in 3.0-rc2

Clock & PM breakage in 3.0-rc2

[Jarkko Nikula]

Hi

While debugging another issue I noticed that McBSP2 clock on Nokia N900
doesn't get disabled in 3.0-rc2 after calling
pm_runtime_put_sync(mcbsp->dev) and the fclk usecount remains active if
the pm_runtime_get_sync(mcbsp->dev) was ever called activating it.

I believe this affects other drivers too as the system is not able to
enter into retention idle even if the mcbsp clocks are off.

I traced breakage to commit 638080c ("OMAP2+ / PM: move runtime
PM implementation to use device power domains").

Reventing that and and 2064af9 ("PM: Revert "driver core: platform_bus:
allow runtime override of dev_pm_ops") makes both mcbsp clocks and
retention idle working again.

Unfortunately I'm not able to see quickly what would be the right fix.

-- 
Jarkko

Re: Clock & PM breakage in 3.0-rc2

[Kevin Hilman]

Jarkko Nikula <jhnikula@gmail.com> writes:

> While debugging another issue I noticed that McBSP2 clock on Nokia N900
> doesn't get disabled in 3.0-rc2 after calling
> pm_runtime_put_sync(mcbsp->dev) and the fclk usecount remains active if
> the pm_runtime_get_sync(mcbsp->dev) was ever called activating it.
>
> I believe this affects other drivers too as the system is not able to
> enter into retention idle even if the mcbsp clocks are off.
>
> I traced breakage to commit 638080c ("OMAP2+ / PM: move runtime
> PM implementation to use device power domains").
>
> Reventing that and and 2064af9 ("PM: Revert "driver core: platform_bus:
> allow runtime override of dev_pm_ops") makes both mcbsp clocks and
> retention idle working again.
>
> Unfortunately I'm not able to see quickly what would be the right fix.

This should be fixed in my for_3.0/pm-fixes branch (pull req. sent to
Tony earlier this week.)

Specifically, the "OMAP: PM: omap_device: fix device power domain callbacks"
patch should fix your problem

Can you test again with merging that branch from:

    git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-omap-pm.git

Kevin

Re: Clock & PM breakage in 3.0-rc2

[Jarkko Nikula]

On Wed, 08 Jun 2011 11:32:47 -0700
Kevin Hilman <khilman@ti.com> wrote:

> > I traced breakage to commit 638080c ("OMAP2+ / PM: move runtime
> > PM implementation to use device power domains").
> >
> > Reventing that and and 2064af9 ("PM: Revert "driver core: platform_bus:
> > allow runtime override of dev_pm_ops") makes both mcbsp clocks and
> > retention idle working again.
> >
> > Unfortunately I'm not able to see quickly what would be the right fix.
> 
> This should be fixed in my for_3.0/pm-fixes branch (pull req. sent to
> Tony earlier this week.)
> 
> Specifically, the "OMAP: PM: omap_device: fix device power domain callbacks"
> patch should fix your problem
> 
> Can you test again with merging that branch from:
> 
>     git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-omap-pm.git
> 
Great! That patch fixes the mcbsp clock problem and I suppose retention
too (I don't have a multimeter with me now to verify but retention
counters in /sys/kernel/debug/pm_debug/count are increasing now).

-- 
Jarkko