Re: [RFC/PATCH 1/2] cpuidle: Allow idle-states to be disabled at start

[Daniel Lezcano]

On 08/25/2016 03:46 PM, Balbir Singh wrote:
> 
> 
> On 25/08/16 01:06, Daniel Lezcano wrote:
>> On 08/24/2016 04:48 PM, Balbir Singh wrote:
>>>
>>>
>>> On 25/08/16 00:44, Daniel Lezcano wrote:
 On 08/19/2016 12:26 AM, Gautham R. Shenoy wrote:
> From: "Gautham R. Shenoy" 
>
> Currently all the idle states registered by a cpu-idle driver are
> enabled by default. This patch adds a mechanism which allows the
> driver to hint if an idle-state should start in a disabled state. The
> cpu-idle core will use this hint to appropriately initialize the
> usage->disable knob of the CPU device idle state.

 Why do you need to do that ?

>>>
>>> I think patch 2/2 explains the reason as it uses this infrastructure
>>
>> Ok, let me elaborate the question, I was not clear.
>>
>> Why the userspace can't setup the system environment at boot time by
>> disabling the state instead of adding extra code to disable it at boot
>> time in the kernel and then re-enable it from userspace ?
> 
> Gautham's patches don't want to have those states enabled by default.
> They are unlikely to be what production systems need, but likely
> what a knowledgeable person can look into selectively enable for
> experimentation.

Why not invert the logic ?

A knowledgeable person can look into selectively disable for production.

In addition, a kernel command line option to specify which state to
disable would be appropriate and beneficial for all existing drivers.


-- 
  Linaro.org │ Open source software for ARM SoCs

Follow Linaro:   Facebook |
 Twitter |
 Blog

Re: [RFC/PATCH 1/2] cpuidle: Allow idle-states to be disabled at start

[Balbir Singh]

On 25/08/16 01:06, Daniel Lezcano wrote:
> On 08/24/2016 04:48 PM, Balbir Singh wrote:
>>
>>
>> On 25/08/16 00:44, Daniel Lezcano wrote:
>>> On 08/19/2016 12:26 AM, Gautham R. Shenoy wrote:
 From: "Gautham R. Shenoy" 

 Currently all the idle states registered by a cpu-idle driver are
 enabled by default. This patch adds a mechanism which allows the
 driver to hint if an idle-state should start in a disabled state. The
 cpu-idle core will use this hint to appropriately initialize the
 usage->disable knob of the CPU device idle state.
>>>
>>> Why do you need to do that ?
>>>
>>
>> I think patch 2/2 explains the reason as it uses this infrastructure
> 
> Ok, let me elaborate the question, I was not clear.
> 
> Why the userspace can't setup the system environment at boot time by
> disabling the state instead of adding extra code to disable it at boot
> time in the kernel and then re-enable it from userspace ?

Gautham's patches don't want to have those states enabled by default.
They are unlikely to be what production systems need, but likely
what a knowledgeable person can look into selectively enable for
experimentation.

@Gautham?


Balbir Singh.

Re: [RFC/PATCH 1/2] cpuidle: Allow idle-states to be disabled at start

[Daniel Lezcano]

On 08/24/2016 04:48 PM, Balbir Singh wrote:
> 
> 
> On 25/08/16 00:44, Daniel Lezcano wrote:
>> On 08/19/2016 12:26 AM, Gautham R. Shenoy wrote:
>>> From: "Gautham R. Shenoy" 
>>>
>>> Currently all the idle states registered by a cpu-idle driver are
>>> enabled by default. This patch adds a mechanism which allows the
>>> driver to hint if an idle-state should start in a disabled state. The
>>> cpu-idle core will use this hint to appropriately initialize the
>>> usage->disable knob of the CPU device idle state.
>>
>> Why do you need to do that ?
>>
> 
> I think patch 2/2 explains the reason as it uses this infrastructure

Ok, let me elaborate the question, I was not clear.

Why the userspace can't setup the system environment at boot time by
disabling the state instead of adding extra code to disable it at boot
time in the kernel and then re-enable it from userspace ?

  -- Daniel


-- 
  Linaro.org │ Open source software for ARM SoCs

Follow Linaro:   Facebook |
 Twitter |
 Blog

Re: [RFC/PATCH 1/2] cpuidle: Allow idle-states to be disabled at start

[Balbir Singh]

On 25/08/16 00:44, Daniel Lezcano wrote:
> On 08/19/2016 12:26 AM, Gautham R. Shenoy wrote:
>> From: "Gautham R. Shenoy" 
>>
>> Currently all the idle states registered by a cpu-idle driver are
>> enabled by default. This patch adds a mechanism which allows the
>> driver to hint if an idle-state should start in a disabled state. The
>> cpu-idle core will use this hint to appropriately initialize the
>> usage->disable knob of the CPU device idle state.
> 
> Why do you need to do that ?
> 

I think patch 2/2 explains the reason as it uses this infrastructure

Balbir Singh

Re: [RFC/PATCH 1/2] cpuidle: Allow idle-states to be disabled at start

[Daniel Lezcano]

On 08/19/2016 12:26 AM, Gautham R. Shenoy wrote:
> From: "Gautham R. Shenoy" 
> 
> Currently all the idle states registered by a cpu-idle driver are
> enabled by default. This patch adds a mechanism which allows the
> driver to hint if an idle-state should start in a disabled state. The
> cpu-idle core will use this hint to appropriately initialize the
> usage->disable knob of the CPU device idle state.

Why do you need to do that ?

> The state can be enabled at run time by echo'ing a zero to the sysfs
> "disable" control file.

... for each cpu.


-- 
  Linaro.org │ Open source software for ARM SoCs

Follow Linaro:   Facebook |
 Twitter |
 Blog