Re: [PATCH v2 2/3] sched: Cleanup SCHED_THERMAL_PRESSURE kconfig entry

[Valentin Schneider <valentin.schneider@arm.com>]

Hi,

On 27/07/20 18:45, Dietmar Eggemann wrote:
> On 27/07/2020 16:18, Qian Cai wrote:
>> On Sun, Jul 12, 2020 at 05:59:16PM +0100, Valentin Schneider wrote:
>>> As Russell pointed out [1], this option is severely lacking in the
>>> documentation department, and figuring out if one has the required
>>> dependencies to benefit from turning it on is not straightforward.
>>>
>>> Make it non user-visible, and add a bit of help to it. While at it, make it
>>> depend on CPU_FREQ_THERMAL.
>>>
>>> [1]: https://lkml.kernel.org/r/20200603173150.GB1551@shell.armlinux.org.uk
>>>
>>> Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
>>> ---
>>>  init/Kconfig | 15 ++++++++++++++-
>>>  1 file changed, 14 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/init/Kconfig b/init/Kconfig
>>> index 0498af567f70..0a97d85568b2 100644
>>> --- a/init/Kconfig
>>> +++ b/init/Kconfig
>>> @@ -492,8 +492,21 @@ config HAVE_SCHED_AVG_IRQ
>>>     depends on SMP
>>>
>>>  config SCHED_THERMAL_PRESSURE
>>> -	bool "Enable periodic averaging of thermal pressure"
>>> +	bool
>>>     depends on SMP
>>> +	depends on CPU_FREQ_THERMAL
>>> +	help
>>> +	  Select this option to enable thermal pressure accounting in the
>>> +	  scheduler. Thermal pressure is the value conveyed to the scheduler
>>> +	  that reflects the reduction in CPU compute capacity resulted from
>>> +	  thermal throttling. Thermal throttling occurs when the performance of
>>> +	  a CPU is capped due to high operating temperatures.
>>> +
>>> +	  If selected, the scheduler will be able to balance tasks accordingly,
>>> +	  i.e. put less load on throttled CPUs than on non/less throttled ones.
>>> +
>>> +	  This requires the architecture to implement
>>> +	  arch_set_thermal_pressure() and arch_get_thermal_pressure().
>>>
>>>  config BSD_PROCESS_ACCT
>>>     bool "BSD Process Accounting"
>>> --
>>
>> On arm64 linux-next (20200727),
>>
>> https://gitlab.com/cailca/linux-mm/-/blob/master/arm64.config
>>
>> WARNING: unmet direct dependencies detected for SCHED_THERMAL_PRESSURE
>>   Depends on [n]: SMP [=y] && CPU_FREQ_THERMAL [=n]
>>   Selected by [y]:
>>   - ARM64 [=y]
>
> Not sure, but:
>
> (1) do we wan to let people enable SCHED_THERMAL_PRESSURE for arm64 so
> arm64 can potentially run w/o a CPU freq cooling device?
>
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index 2d4abbc9f8d0..baffe8b66da2 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -192,7 +192,6 @@ config ARM64
>         select PCI_SYSCALL if PCI
>         select POWER_RESET
>         select POWER_SUPPLY
> -       select SCHED_THERMAL_PRESSURE
>         select SPARSE_IRQ
>         select SWIOTLB
>         select SYSCTL_EXCEPTION_TRACE
> diff --git a/init/Kconfig b/init/Kconfig
> index 37b089f87804..8b36e07fb230 100644
> --- a/init/Kconfig
> +++ b/init/Kconfig
> @@ -492,7 +492,7 @@ config HAVE_SCHED_AVG_IRQ
>         depends on SMP
>
>  config SCHED_THERMAL_PRESSURE
> -       bool
> +       bool "Thermal pressure accounting"
>         depends on SMP
>         depends on CPU_FREQ_THERMAL
>         help
>
> Or
>
> (2) should SCHED_THERMAL_PRESSURE for arm64 be enabled by default?
>
> But then it makes no sense to allow the removal of CONFIG_CPU_FREQ_THERMAL.
>
> linux-next/master$ make ARCH=arm64 defconfig
>
> // Remove CONFIG_CPU_FREQ_THERMAL
> linux-next/master$ grep CPU_FREQ_THERMAL .config
> # CONFIG_CPU_FREQ_THERMAL is not set
>
> linux-next/master$ make
> scripts/kconfig/conf  --syncconfig Kconfig
>
> WARNING: unmet direct dependencies detected for SCHED_THERMAL_PRESSURE
>   Depends on [n]: SMP [=y] && CPU_FREQ_THERMAL [=n]
>   Selected by [y]:
>   - ARM64 [=y]
>
> WARNING: unmet direct dependencies detected for SCHED_THERMAL_PRESSURE
>   Depends on [n]: SMP [=y] && CPU_FREQ_THERMAL [=n]
>   Selected by [y]:
>   - ARM64 [=y]
>
> WARNING: unmet direct dependencies detected for SCHED_THERMAL_PRESSURE
>   Depends on [n]: SMP [=y] && CPU_FREQ_THERMAL [=n]
>   Selected by [y]:
>   - ARM64 [=y]
>   HOSTCC  scripts/dtc/dtc.o
>
> ---
>
> There is a similar issue with arm.
>
> I would prefer for (1).

I went for having SCHED_THERMAL_PRESSURE in arm64/Kconfig because of where
the discussion went in the original thread ([1] in the changelog).

One point is that selecting this option requires having the right
infrastructure in place (arch_{set, scale}_thermal_pressure() must be
redefined by the architecture), which cannot be easily expressed in Kconfig
terms. Russell's point was that this is difficult for a lambda user to make
sense of, and Vincent argued that this option should simply be selected at
architecture level, which, given the context, makes sense IMO.

We could change the arch Kconfig into

  select SCHED_THERMAL_PRESSURE if CPU_FREQ_THERMAL

but that seems redundant; this dependency is already expressed in
SCHED_THERMAL_PRESSURE's definition. Is there a proper pattern to select
some Kconfig option only if all of its dependencies are met?