On Fri, Dec 13, 2013 at 01:56:18PM -0600, Felipe Balbi wrote:
> On Fri, Dec 13, 2013 at 02:01:32PM +0900, Anton Tikhomirov wrote:
> > Hi Felipe,
> > 
> > > -static int dwc3_exynos_suspend(struct device *dev)
> > > +static int __dwc3_exynos_suspend(struct dwc3_exynos *exynos)
> > >  {
> > > -	struct dwc3_exynos *exynos = dev_get_drvdata(dev);
> > > -
> > >  	clk_disable(exynos->clk);
> > > 
> > >  	return 0;
> > >  }
> > > 
> > > +static int __dwc3_exynos_resume(struct dwc3_exynos *exynos)
> > > +{
> > > +	return clk_enable(exynos->clk);
> > > +}
> > > +
> > > +static int dwc3_exynos_suspend(struct device *dev)
> > > +{
> > > +	struct dwc3_exynos *exynos = dev_get_drvdata(dev);
> > > +
> > > +	return __dwc3_exynos_suspend(exynos);
> > 
> > If dwc3-exynos is runtime suspended, the clock will be disabled
> > second time here (unbalanced clk_enable/clk_disable).
> 
> I don't get what you mean but there is something that probably needs
> fixing, I guess below makes it better:
> 
> diff --git a/drivers/usb/dwc3/dwc3-exynos.c b/drivers/usb/dwc3/dwc3-exynos.c
> index c93919a..1e5720a 100644
> --- a/drivers/usb/dwc3/dwc3-exynos.c
> +++ b/drivers/usb/dwc3/dwc3-exynos.c
> @@ -218,6 +218,9 @@ static int dwc3_exynos_suspend(struct device *dev)
>  {
>  	struct dwc3_exynos *exynos = dev_get_drvdata(dev);
>  
> +	if (pm_runtime_suspended(dev))
> +		return 0;
> +
>  	return __dwc3_exynos_suspend(exynos);
>  }
>  
> 
> Is that what you meant ?

note, however, that this is *not* a case where we would fall today. See
that we pm_runtime_get() in probe and only pm_runtime_put() during
remove. So there would never be a case where we would try system suspend
while device was already runtime suspended.

I have fixed all patches in my testing/next branch anyway, just to make
sure we're "idiot-proof" when it comes to implementing real runtime pm
later on :-)

cheers

-- 
balbi