[PATCH V2 0/2] Suspend clk scaling when there is no request

[PATCH V2 0/2] Suspend clk scaling when there is no request

[Ram Prakash Gupta]

Currently ufs clk scaling is getting suspended only when the
clks are scaled down, but next when high load is generated its
adding a huge amount of latency in scaling up the clk and complete
the request post that.

Now if the scaling is suspended in its existing state, and when high
load is generated it is helping improve the random performance KPI by
28%. So suspending the scaling when there is no request. And the clk
would be put in low scaled state when the actual request load is low.

Making this change as optional for other vendor by having the check
enabled using vops as for some vendor suspending without bringing the
clk in low scaled state might have impact on power consumption on the
SoC.

Changes since V1:
	- Address minor review comment.

Ram Prakash Gupta (2):
  scsi: ufs: Suspend clk scaling on no request
  scsi: ufs: qcom: Enable suspending clk scaling on no request

 drivers/ufs/core/ufshcd.c   | 3 ++-
 drivers/ufs/host/ufs-qcom.c | 2 ++
 include/ufs/ufshcd.h        | 1 +
 3 files changed, 5 insertions(+), 1 deletion(-)

-- 
2.17.1

[PATCH v2 1/2] scsi: ufs: Suspend clk scaling on no request

[Ram Prakash Gupta]

Currently ufs clk scaling is getting suspended only when the
clks are scaled down, but next when high load is generated its
adding a huge amount of latency in scaling up the clk and complete
the request post that.

Now if the scaling is suspended in its existing state, and when high
load is generated it is helping improve the random performance KPI by
28%. So suspending the scaling when there is no request. And the clk
would be put in low scaled state when the actual request load is low.

Making this change as optional for other vendor by having the check
enabled using vops as for some vendor suspending without bringing the
clk in low scaled state might have impact on power consumption on the
SoC.

Signed-off-by: Ram Prakash Gupta <quic_rampraka@quicinc.com>
---
 drivers/ufs/core/ufshcd.c | 3 ++-
 include/ufs/ufshcd.h      | 1 +
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index 1b65e6ae4137..9f935e5c60e8 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -1560,7 +1560,8 @@ static int ufshcd_devfreq_target(struct device *dev,
 		ktime_to_us(ktime_sub(ktime_get(), start)), ret);
 
 out:
-	if (sched_clk_scaling_suspend_work && !scale_up)
+	if (sched_clk_scaling_suspend_work &&
+			(!scale_up || hba->clk_scaling.suspend_on_no_request))
 		queue_work(hba->clk_scaling.workq,
 			   &hba->clk_scaling.suspend_work);
 
diff --git a/include/ufs/ufshcd.h b/include/ufs/ufshcd.h
index bad88bd91995..c14607f2890b 100644
--- a/include/ufs/ufshcd.h
+++ b/include/ufs/ufshcd.h
@@ -457,6 +457,7 @@ struct ufs_clk_scaling {
 	bool is_initialized;
 	bool is_busy_started;
 	bool is_suspended;
+	bool suspend_on_no_request;
 };
 
 #define UFS_EVENT_HIST_LENGTH 8
-- 
2.17.1

Re: [PATCH v2 1/2] scsi: ufs: Suspend clk scaling on no request

[Bart Van Assche]

On 6/27/24 1:37 AM, Ram Prakash Gupta wrote:
> diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> index 1b65e6ae4137..9f935e5c60e8 100644
> --- a/drivers/ufs/core/ufshcd.c
> +++ b/drivers/ufs/core/ufshcd.c
> @@ -1560,7 +1560,8 @@ static int ufshcd_devfreq_target(struct device *dev,
>   		ktime_to_us(ktime_sub(ktime_get(), start)), ret);
>   
>   out:
> -	if (sched_clk_scaling_suspend_work && !scale_up)
> +	if (sched_clk_scaling_suspend_work &&
> +			(!scale_up || hba->clk_scaling.suspend_on_no_request))
>   		queue_work(hba->clk_scaling.workq,
>   			   &hba->clk_scaling.suspend_work);
>   
> diff --git a/include/ufs/ufshcd.h b/include/ufs/ufshcd.h
> index bad88bd91995..c14607f2890b 100644
> --- a/include/ufs/ufshcd.h
> +++ b/include/ufs/ufshcd.h
> @@ -457,6 +457,7 @@ struct ufs_clk_scaling {
>   	bool is_initialized;
>   	bool is_busy_started;
>   	bool is_suspended;
> +	bool suspend_on_no_request;
>   };
>   
>   #define UFS_EVENT_HIST_LENGTH 8

Who are the other vendors that support clock scaling? I'm asking because
I don't think that the behavior change introduced by this patch should
depend on the SoC vendor.

Thanks,

Bart.

Re: [PATCH v2 1/2] scsi: ufs: Suspend clk scaling on no request

[Ram Prakash Gupta]

On 6/27/2024 10:05 PM, Bart Van Assche wrote:
> On 6/27/24 1:37 AM, Ram Prakash Gupta wrote:
>> diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
>> index 1b65e6ae4137..9f935e5c60e8 100644
>> --- a/drivers/ufs/core/ufshcd.c
>> +++ b/drivers/ufs/core/ufshcd.c
>> @@ -1560,7 +1560,8 @@ static int ufshcd_devfreq_target(struct device 
>> *dev,
>>           ktime_to_us(ktime_sub(ktime_get(), start)), ret);
>>   out:
>> -    if (sched_clk_scaling_suspend_work && !scale_up)
>> +    if (sched_clk_scaling_suspend_work &&
>> +            (!scale_up || hba->clk_scaling.suspend_on_no_request))
>>           queue_work(hba->clk_scaling.workq,
>>                  &hba->clk_scaling.suspend_work);
>> diff --git a/include/ufs/ufshcd.h b/include/ufs/ufshcd.h
>> index bad88bd91995..c14607f2890b 100644
>> --- a/include/ufs/ufshcd.h
>> +++ b/include/ufs/ufshcd.h
>> @@ -457,6 +457,7 @@ struct ufs_clk_scaling {
>>       bool is_initialized;
>>       bool is_busy_started;
>>       bool is_suspended;
>> +    bool suspend_on_no_request;
>>   };
>>   #define UFS_EVENT_HIST_LENGTH 8
> 
> Who are the other vendors that support clock scaling? I'm asking because
> I don't think that the behavior change introduced by this patch should
> depend on the SoC vendor.
> 
> Thanks,
> 
> Bart.

Hi Bart,

I guess, Mediatek is one vendor who is having this feature in use as I
see some fixes coming from Peter with respect to clk scaling, where some
power regression on mediatek chipsets were addressed.

Please check below link for background of this change and reason to keep
it vendor specific.
https://www.spinics.net/lists/linux-scsi/msg193591.html

Thanks,
Ram

[PATCH v2 2/2] scsi: ufs: qcom: Enable suspending clk scaling on no request

[Ram Prakash Gupta]

Enable suspending clk scaling on no request for Qualcomm SoC.

Signed-off-by: Ram Prakash Gupta <quic_rampraka@quicinc.com>
---
 drivers/ufs/host/ufs-qcom.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/ufs/host/ufs-qcom.c b/drivers/ufs/host/ufs-qcom.c
index cca190d1c577..9041ffab152c 100644
--- a/drivers/ufs/host/ufs-qcom.c
+++ b/drivers/ufs/host/ufs-qcom.c
@@ -1548,6 +1548,8 @@ static void ufs_qcom_config_scaling_param(struct ufs_hba *hba,
 	p->timer = DEVFREQ_TIMER_DELAYED;
 	d->upthreshold = 70;
 	d->downdifferential = 5;
+
+	hba->clk_scaling.suspend_on_no_request = true;
 }
 #else
 static void ufs_qcom_config_scaling_param(struct ufs_hba *hba,
-- 
2.17.1

Re: [PATCH V2 0/2] Suspend clk scaling when there is no request

[Bart Van Assche]

On 6/27/24 1:37 AM, Ram Prakash Gupta wrote:
> Changes since V1:
> 	- Address minor review comment.

This is too vague. Please be more specific in patch changelogs.

Thanks,

Bart.

Re: [PATCH V2 0/2] Suspend clk scaling when there is no request

[Ram Prakash Gupta]

On 6/27/2024 10:06 PM, Bart Van Assche wrote:
> On 6/27/24 1:37 AM, Ram Prakash Gupta wrote:
>> Changes since V1:
>>     - Address minor review comment.
> 
> This is too vague. Please be more specific in patch changelogs.
> 
> Thanks,
> 
> Bart.
> 
My Bad - will update it properly if next patchset is required.
By the way here I addressed code review comment from Dmitry.

https://www.spinics.net/lists/kernel/msg5268621.html

Thanks,
Ram

Re: [PATCH V2 0/2] Suspend clk scaling when there is no request

[Bart Van Assche]

On 6/27/24 1:37 AM, Ram Prakash Gupta wrote:
> Currently ufs clk scaling is getting suspended only when the
> clks are scaled down, but next when high load is generated its
> adding a huge amount of latency in scaling up the clk and complete
> the request post that.
> 
> Now if the scaling is suspended in its existing state, and when high
> load is generated it is helping improve the random performance KPI by
> 28%. So suspending the scaling when there is no request. And the clk
> would be put in low scaled state when the actual request load is low.
> 
> Making this change as optional for other vendor by having the check
> enabled using vops as for some vendor suspending without bringing the
> clk in low scaled state might have impact on power consumption on the
> SoC.

For both patches:

Reviewed-by: Bart Van Assche <bvanassche@acm.org>

Re: [PATCH V2 0/2] Suspend clk scaling when there is no request

[Martin K. Petersen]

Ram,

> Currently ufs clk scaling is getting suspended only when the clks are
> scaled down, but next when high load is generated its adding a huge
> amount of latency in scaling up the clk and complete the request post
> that.

Applied to 6.11/scsi-staging, thanks!

-- 
Martin K. Petersen	Oracle Linux Engineering

Re: [PATCH V2 0/2] Suspend clk scaling when there is no request

[Martin K. Petersen]

On Thu, 27 Jun 2024 14:07:54 +0530, Ram Prakash Gupta wrote:

> Currently ufs clk scaling is getting suspended only when the
> clks are scaled down, but next when high load is generated its
> adding a huge amount of latency in scaling up the clk and complete
> the request post that.
> 
> Now if the scaling is suspended in its existing state, and when high
> load is generated it is helping improve the random performance KPI by
> 28%. So suspending the scaling when there is no request. And the clk
> would be put in low scaled state when the actual request load is low.
> 
> [...]

Applied to 6.11/scsi-queue, thanks!

[1/2] scsi: ufs: Suspend clk scaling on no request
      https://git.kernel.org/mkp/scsi/c/50183ac2cfb5
[2/2] scsi: ufs: qcom: Enable suspending clk scaling on no request
      https://git.kernel.org/mkp/scsi/c/ed7dac86f140

-- 
Martin K. Petersen	Oracle Linux Engineering