[PATCH] ARM: kernel: document ARM CPUs clock-frequency property

[PATCH] ARM: kernel: document ARM CPUs clock-frequency property

[Florian Fainelli]

ARM CPU device tree nodes may contain an optional clock-frequency
property, when set, this property must contain the CPU frequency in Hz,
which is then used by the topology parsing code in
arch/arm/kernel/topology.c to infer the CPU capacity.

Document this property to avoid any possible confusion on the
clock-frequency unit.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 Documentation/devicetree/bindings/arm/cpus.txt | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/Documentation/devicetree/bindings/arm/cpus.txt b/Documentation/devicetree/bindings/arm/cpus.txt
index f32494d..a33b956 100644
--- a/Documentation/devicetree/bindings/arm/cpus.txt
+++ b/Documentation/devicetree/bindings/arm/cpus.txt
@@ -45,6 +45,8 @@ For the ARM architecture every CPU node must contain the following properties:
 		"marvell,xsc3"
 		"marvell,xscale"
 
+- clock-frequency : The frequency of the CPU, in Hz. Optional.
+
 Example:
 
 	cpus {
-- 
1.8.1.2

[PATCH] ARM: kernel: document ARM CPUs clock-frequency property

[Mark Rutland]

On Wed, Jun 19, 2013 at 10:47:17AM +0100, Florian Fainelli wrote:
> ARM CPU device tree nodes may contain an optional clock-frequency
> property, when set, this property must contain the CPU frequency in Hz,
> which is then used by the topology parsing code in
> arch/arm/kernel/topology.c to infer the CPU capacity.

I see that arch/arm/kernel/topology.c does a pr_err when clock-frequency isn't
present. If we're documenting it as optional, should we not downgrade that or
remove it entirely?

Thanks,
Mark.

> 
> Document this property to avoid any possible confusion on the
> clock-frequency unit.
> 
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
> ---
>  Documentation/devicetree/bindings/arm/cpus.txt | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/Documentation/devicetree/bindings/arm/cpus.txt b/Documentation/devicetree/bindings/arm/cpus.txt
> index f32494d..a33b956 100644
> --- a/Documentation/devicetree/bindings/arm/cpus.txt
> +++ b/Documentation/devicetree/bindings/arm/cpus.txt
> @@ -45,6 +45,8 @@ For the ARM architecture every CPU node must contain the following properties:
>  		"marvell,xsc3"
>  		"marvell,xscale"
>  
> +- clock-frequency : The frequency of the CPU, in Hz. Optional.
> +
>  Example:
>  
>  	cpus {
> -- 
> 1.8.1.2
> 
> 
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
> 

[PATCH] ARM: kernel: document ARM CPUs clock-frequency property

[Florian Fainelli]

Hello Mark,

2013/6/19 Mark Rutland <mark.rutland@arm.com>:
> On Wed, Jun 19, 2013 at 10:47:17AM +0100, Florian Fainelli wrote:
>> ARM CPU device tree nodes may contain an optional clock-frequency
>> property, when set, this property must contain the CPU frequency in Hz,
>> which is then used by the topology parsing code in
>> arch/arm/kernel/topology.c to infer the CPU capacity.
>
> I see that arch/arm/kernel/topology.c does a pr_err when clock-frequency isn't
> present. If we're documenting it as optional, should we not downgrade that or
> remove it entirely?

Good question. I think it might be good to just dowgrade the error to
a simple warning, we should clearly be fixing the binding if we have a
device_type = "cpu" property but no clock-frequency property.
--
Florian

[PATCH] ARM: kernel: document ARM CPUs clock-frequency property

[Lorenzo Pieralisi]

On Wed, Jun 19, 2013 at 11:27:50AM +0100, Florian Fainelli wrote:
> Hello Mark,
> 
> 2013/6/19 Mark Rutland <mark.rutland@arm.com>:
> > On Wed, Jun 19, 2013 at 10:47:17AM +0100, Florian Fainelli wrote:
> >> ARM CPU device tree nodes may contain an optional clock-frequency
> >> property, when set, this property must contain the CPU frequency in Hz,
> >> which is then used by the topology parsing code in
> >> arch/arm/kernel/topology.c to infer the CPU capacity.
> >
> > I see that arch/arm/kernel/topology.c does a pr_err when clock-frequency isn't
> > present. If we're documenting it as optional, should we not downgrade that or
> > remove it entirely?
> 
> Good question. I think it might be good to just dowgrade the error to
> a simple warning, we should clearly be fixing the binding if we have a
> device_type = "cpu" property but no clock-frequency property.

Well, this means that that pr_err will trigger as soon as the dts
updates hit the mainline, now probably it is silent because cpu nodes
for architectures with "arm,cortex-a15" as cpus miss device_type = "cpu".
ePAPR defines "clock-frequency" as required.

So either we downgrade it to optional or we are in for another slew of dts
updates, I am really looking forward to that. Certainly it has to be added
to the bindings.

Thoughts ?

Lorenzo

[PATCH] ARM: kernel: document ARM CPUs clock-frequency property

[Florian Fainelli]

2013/6/19 Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>:
> On Wed, Jun 19, 2013 at 11:27:50AM +0100, Florian Fainelli wrote:
>> Hello Mark,
>>
>> 2013/6/19 Mark Rutland <mark.rutland@arm.com>:
>> > On Wed, Jun 19, 2013 at 10:47:17AM +0100, Florian Fainelli wrote:
>> >> ARM CPU device tree nodes may contain an optional clock-frequency
>> >> property, when set, this property must contain the CPU frequency in Hz,
>> >> which is then used by the topology parsing code in
>> >> arch/arm/kernel/topology.c to infer the CPU capacity.
>> >
>> > I see that arch/arm/kernel/topology.c does a pr_err when clock-frequency isn't
>> > present. If we're documenting it as optional, should we not downgrade that or
>> > remove it entirely?
>>
>> Good question. I think it might be good to just dowgrade the error to
>> a simple warning, we should clearly be fixing the binding if we have a
>> device_type = "cpu" property but no clock-frequency property.
>
> Well, this means that that pr_err will trigger as soon as the dts
> updates hit the mainline, now probably it is silent because cpu nodes
> for architectures with "arm,cortex-a15" as cpus miss device_type = "cpu".
> ePAPR defines "clock-frequency" as required.

Correct. I am also fine with documenting this property as mandatory.

>
> So either we downgrade it to optional or we are in for another slew of dts
> updates, I am really looking forward to that. Certainly it has to be added
> to the bindings.
>
> Thoughts ?

Well, the fix would be twofold:
- for platforms which already have an "arm,cortex-a15" CPU nodes but
lack a clock phandle or a clock-frequency add it
- update arch/arm/kernel/toplogy.c to fetch the clock phandle and rate
if specified (e.g: ecx-2000)

Here is a list of platforms/DTS that would appear to need fixing:
- exynos5440.dtsi
- tegra114.dtsi
- vexpress-v2p-ca15-tc1.dts
- vexpress-v2p-ca15_a7.dts
- xenvm-4.2.dts

As far as I am concerned, the DTB passed to my kernel is already okay,
but certainly this needs fixing for in-tree DTS files.
--
Florian