linux-kernel - Re: [PATCH] thermal: power_allocate: add upper and lower limits

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <accb83e0-ffbe-b6e3-6bf9-e7cc8b9fe19c@arm.com>
Date:   Fri, 24 Apr 2020 10:22:03 +0100
From:   Lukasz Luba <lukasz.luba@....com>
To:     Michael Kao <michael.kao@...iatek.com>,
        Zhang Rui <rui.zhang@...el.com>,
        Eduardo Valentin <edubezval@...il.com>,
        Daniel Lezcano <daniel.lezcano@...aro.org>,
        Rob Herring <robh+dt@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Matthias Brugger <matthias.bgg@...il.com>, hsinyi@...omium.org,
        linux-pm@...r.kernel.org, srv_heupstream@...iatek.com
Cc:     devicetree@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org,
        linux-mediatek@...ts.infradead.org
Subject: Re: [PATCH] thermal: power_allocate: add upper and lower limits

Hi Michael,

On 4/24/20 8:16 AM, Michael Kao wrote:
> The upper and lower limits of thermal throttle state in the
> device tree do not apply to the power_allocate governor.
> Add the upper and lower limits to the power_allocate governor.
> 
> Signed-off-by: Michael Kao <michael.kao@...iatek.com>
> ---
>   drivers/thermal/thermal_core.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/thermal/thermal_core.c b/drivers/thermal/thermal_core.c
> index 9a321dc548c8..f6feed2265bd 100644
> --- a/drivers/thermal/thermal_core.c
> +++ b/drivers/thermal/thermal_core.c
> @@ -598,7 +598,7 @@ int power_actor_set_power(struct thermal_cooling_device *cdev,
>   	if (ret)
>   		return ret;
>   
> -	instance->target = state;
> +	instance->target = clamp_val(state, instance->lower, instance->upper);
>   	mutex_lock(&cdev->lock);
>   	cdev->updated = false;
>   	mutex_unlock(&cdev->lock);
> 

Thank you for the patch and having to look at it. I have some concerns
with this approach. Let's analyze it further.

In default the cooling devices in the thermal zone which is used by IPA
do not have this 'lower' and 'upper' limits. They are set to
THERMAL_NO_LIMIT in DT to give full control to IPA over the states.

This the function 'power_actor_set_power' actually translates granted
power to the state that device will run for the next period.
The IPA algorithm has already split the power budget.
Now what happen when the 'lower' value will change the state to a state
which consumes more power than was calculated in the IPA alg... It will
became unstable.

I would rather see a change which uses these 'lower' and 'upper' limits
before the IPA do the calculation of the power budget. But this wasn't
a requirement and we assumed that IPA has full control over the cooling
device (which I described above with this DT THERMAL_NO_LIMIT).

Is there a problem with your platform that it has to provide some
minimal performance, so you tried to introduce this clamping?

Regards,
Lukasz