RE: [RFC PATCH] cpuidle: Make cpuidle governor switchable to be the default behaviour

["Doug Smythies" <dsmythies@telus.net>]

I very much support this RFC.
I have been running only with "cpuidle_sysfs_switch" for about 2 years.

Some changes would be required for the documentation files also.

On 2020.04.27 06:37 Daniel Lezcano wrote:
> On 27/04/2020 12:17, Hanjun Guo wrote:
>> For now cpuidle governor can be switched via sysfs only when the
>> boot option "cpuidle_sysfs_switch" is passed, but it's important
>> to switch the governor to adapt to different workloads, especially
>> after TEO and haltpoll governor were introduced.
>> 
>> Introduce a CONFIG option to make cpuidle governor switchable to be
>> the default behaviour, which will not break the boot option behaviour.
>> 
>> Signed-off-by: Hanjun Guo <guohanjun@huawei.com>
>> ---
>>  drivers/cpuidle/Kconfig | 9 +++++++++
>>  drivers/cpuidle/sysfs.c | 2 +-
>>  2 files changed, 10 insertions(+), 1 deletion(-)
>> 
>> diff --git a/drivers/cpuidle/Kconfig b/drivers/cpuidle/Kconfig
>> index c0aeedd..c40cb40 100644
>> --- a/drivers/cpuidle/Kconfig
>> +++ b/drivers/cpuidle/Kconfig
>> @@ -47,6 +47,15 @@ config CPU_IDLE_GOV_HALTPOLL
>>  config DT_IDLE_STATES
>>  	bool
>>  
>> +config CPU_IDLE_SWITCH_GOV_IN_DEFAULT
>> +	bool "Switch the CPU idle governor via sysfs at runtime in default behaviour"
>> +	help
>> +	  Make the CPU idle governor switchable at runtime, and make it as the
>> +	  default behaviour even the boot option "cpuidle_sysfs_switch" is not
>> +	  passed in cmdline.
>> +
>> +	  Say N if you unsure about this.
>
> Well I wouldn't make this optional but just remove the sysfs_switch.

Agree.

> However, there is the '_ro' suffix when the option is not set. In order
> to not break the existing tools, may be let both files co-exist and add
> in the ABI/obselete the '_ro' file as candidate for removal ?

I do not like this _ro thing, and got hit by it with turbostat one time.
Agree it should be made a candidate for removal.