[RFC] rk3368 idle states

[RFC] rk3368 idle states

[Lorenzo Pieralisi]

Hi Heiko,

while reviewing other dts idle states bindings I spotted the following
snippet in arch/arm64/boot/dts/rockchip/rk3368.dtsi:

cpu_sleep: cpu-sleep-0 {
       compatible = "arm,idle-state";
       arm,psci-suspend-param = <0x1010000>;
       entry-latency-us = <0x3fffffff>;
       exit-latency-us = <0x40000000>;
       min-residency-us = <0xffffffff>;
};

Could you please explain to me how you obtained these residency/latency
values ? They make NO sense whatsoever and I am quite tempted to
ask you to remove this data from the dts file, it is just wrong.

Thanks,
Lorenzo

[RFC] rk3368 idle states

[Heiko Stübner]

Hi Lorenzo,

Am Mittwoch, 2. Dezember 2015, 11:18:58 schrieb Lorenzo Pieralisi:
> while reviewing other dts idle states bindings I spotted the following
> snippet in arch/arm64/boot/dts/rockchip/rk3368.dtsi:
> 
> cpu_sleep: cpu-sleep-0 {
>        compatible = "arm,idle-state";
>        arm,psci-suspend-param = <0x1010000>;
>        entry-latency-us = <0x3fffffff>;
>        exit-latency-us = <0x40000000>;
>        min-residency-us = <0xffffffff>;
> };
> 
> Could you please explain to me how you obtained these residency/latency
> values ? They make NO sense whatsoever and I am quite tempted to
> ask you to remove this data from the dts file, it is just wrong.

I took these values verbatim from the vendor kernel when adding the core 
rk3368 support.

I guess I was overly enthusiastic, that psci worked out of the box for smp, 
that I didn't check the latency/residency values enough at that time. As time-
values they are really overly long :-) .


Nevertheless I have inquired now, what the background of those values is.


Heiko

[RFC] rk3368 idle states

[Lorenzo Pieralisi]

Hi Heiko,

On Thu, Dec 03, 2015 at 02:05:01AM +0100, Heiko St?bner wrote:
> Hi Lorenzo,
> 
> Am Mittwoch, 2. Dezember 2015, 11:18:58 schrieb Lorenzo Pieralisi:
> > while reviewing other dts idle states bindings I spotted the following
> > snippet in arch/arm64/boot/dts/rockchip/rk3368.dtsi:
> > 
> > cpu_sleep: cpu-sleep-0 {
> >        compatible = "arm,idle-state";
> >        arm,psci-suspend-param = <0x1010000>;
> >        entry-latency-us = <0x3fffffff>;
> >        exit-latency-us = <0x40000000>;
> >        min-residency-us = <0xffffffff>;
> > };
> > 
> > Could you please explain to me how you obtained these residency/latency
> > values ? They make NO sense whatsoever and I am quite tempted to
> > ask you to remove this data from the dts file, it is just wrong.
> 
> I took these values verbatim from the vendor kernel when adding the core 
> rk3368 support.
> 
> I guess I was overly enthusiastic, that psci worked out of the box for smp, 
> that I didn't check the latency/residency values enough at that time. As time-
> values they are really overly long :-) .
> 
> 
> Nevertheless I have inquired now, what the background of those values is.

Thank you, they must be updated, hopefully they won't unearth any issue
with the idle state - ie at present the idle state simply turns out unused.

Lorenzo