Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 20 Jul 2023 13:56:02 +0530
From:   Manivannan Sadhasivam <mani@kernel.org>
To:     Nitin Rawat <quic_nitirawa@quicinc.com>
Cc:     quic_cang@quicinc.com, stanley.chu@mediatek.com,
        bvanassche@acm.org, quic_asutoshd@quicinc.com, avri.altman@wdc.com,
        martin.petersen@oracle.com, beanhuo@micron.com,
        konrad.dybcio@linaro.org, linux-scsi@vger.kernel.org,
        linux-kernel@vger.kernel.org, agross@kernel.org,
        andersson@kernel.org, jejb@linux.ibm.com,
        linux-arm-msm@vger.kernel.org, quic_ziqichen@quicinc.com
Subject: Re: [PATCH V2] scsi: ufs: ufs-qcom: Update UFS devfreq Parameters
Message-ID: <20230720082602.GA4863@thinkpad>
References: <20230716074907.12356-1-quic_nitirawa@quicinc.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20230716074907.12356-1-quic_nitirawa@quicinc.com>
Precedence: bulk

On Sun, Jul 16, 2023 at 01:19:07PM +0530, Nitin Rawat wrote:
> To support the periodic polling mode without stop caused by CPU idle
> state, enable delayed timer as default instead of deferrable timer
> for qualcomm platforms.

How about,

"Devfreq uses the default DEVFREQ_TIMER_DEFERRABLE mode which uses the deferred
timer for scheduling the devfreq load monitor function. This causes the load
monitoring to be done only with non-idle CPUs and not making use of the idle
CPUs. Hence, use the DEVFREQ_TIMER_DELAYED mode which uses the delayed timer
thereby making use of idle CPUs as well for load monitoring."

> And change UFS devfreq downdifferential threshold to 65 for less
> aggressive downscaling.
> 

This change needs to be done in a separate patch. Also please provide any
benchmark if available to justify the patch.

> Signed-off-by: Nitin Rawat <quic_nitirawa@quicinc.com>
> Signed-off-by: Asutosh Das <quic_asutoshd@quicinc.com>

Your signed-off-by should come last, as you are the one sending this patch.

- Mani

> ---
> 
> Changes in v2:
> - Realigned the commit text
> 
>  drivers/ufs/host/ufs-qcom.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/ufs/host/ufs-qcom.c b/drivers/ufs/host/ufs-qcom.c
> index 82d02e7f3b4f..a15815c951ca 100644
> --- a/drivers/ufs/host/ufs-qcom.c
> +++ b/drivers/ufs/host/ufs-qcom.c
> @@ -1388,8 +1388,9 @@ static void ufs_qcom_config_scaling_param(struct ufs_hba *hba,
>  					struct devfreq_simple_ondemand_data *d)
>  {
>  	p->polling_ms = 60;
> +	p->timer = DEVFREQ_TIMER_DELAYED;
>  	d->upthreshold = 70;
> -	d->downdifferential = 5;
> +	d->downdifferential = 65;
>  }
>  #else
>  static void ufs_qcom_config_scaling_param(struct ufs_hba *hba,
> --
> 2.17.1
> 

-- 
மணிவண்ணன் சதாசிவம்