[PATCH] ARM: OMAP4: PM: Avoid expensive cpu_suspend() path for all CPU power states except off

[Kevin Hilman]

Santosh Shilimkar <santosh.shilimkar at ti.com> writes:

> Current CPU PM code code make use of common cpu_suspend() path for all the
> CPU power states which is not optimal. In fact cpu_suspend() path is needed
> only when we put CPU power domain to off state where the CPU context is lost.
>
> Update the code accordingly so that the expensive cpu_suspend() path
> can be avoided for the shallow CPU power states like CPU PD INA/CSWR.
>
> Cc: Kevin Hilman <khilman at deeprootsystems.com>
>
> Reported-by: Richard Woodruff <r-woodruff2 at ti.com>
> Signed-off-by: Santosh Shilimkar <santosh.shilimkar at ti.com>

Looks OK at first glance, but are you sure this is right for the
various ways both clusters might idle using coupled CPUidle?

Some description of the testing would be helpful here.

Thanks,

Kevin

> ---
>  arch/arm/mach-omap2/omap-mpuss-lowpower.c |    5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/arch/arm/mach-omap2/omap-mpuss-lowpower.c b/arch/arm/mach-omap2/omap-mpuss-lowpower.c
> index aac46bf..abdd0f6 100644
> --- a/arch/arm/mach-omap2/omap-mpuss-lowpower.c
> +++ b/arch/arm/mach-omap2/omap-mpuss-lowpower.c
> @@ -276,7 +276,10 @@ int omap4_enter_lowpower(unsigned int cpu, unsigned int power_state)
>  	/*
>  	 * Call low level function  with targeted low power state.
>  	 */
> -	cpu_suspend(save_state, omap4_finish_suspend);
> +	if (save_state)
> +		cpu_suspend(save_state, omap4_finish_suspend);
> +	else
> +		omap4_finish_suspend(save_state);
>  
>  	/*
>  	 * Restore the CPUx power state to ON otherwise CPUx