diff mbox series

[PATCHv2,6/6] thermal: exynos: Add runtime power management for tmu

Message ID 20220515064126.1424-7-linux.amoon@gmail.com
State New
Headers show
Series Exynos Thermal code inprovement | expand

Commit Message

Anand Moon May 15, 2022, 6:41 a.m. UTC
Add runtime power management for exynos thermal driver.

Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Signed-off-by: Anand Moon <linux.amoon@gmail.com>
---
v1: new patch in this series.
---
 drivers/thermal/samsung/exynos_tmu.c | 29 ++++++++++++++++++++++++++--
 1 file changed, 27 insertions(+), 2 deletions(-)

Comments

Anand Moon May 21, 2022, 9:52 a.m. UTC | #1
Hi Krzysztof,

On Wed, 18 May 2022 at 12:49, Krzysztof Kozlowski
<krzysztof.kozlowski@linaro.org> wrote:
>
> On 17/05/2022 20:45, Anand Moon wrote:
> > Hi Krzysztof,
> >
> > On Sun, 15 May 2022 at 15:18, Krzysztof Kozlowski
> > <krzysztof.kozlowski@linaro.org> wrote:
> >>
> >> On 15/05/2022 08:41, Anand Moon wrote:
> >>> Add runtime power management for exynos thermal driver.
> >>
> >> First of all - why? Second, I do not see it being added. Where are the
> >> runtime callbacks?
> >>
> >
> > To control runtime control PMU, did I miss something?
>
> Controlling runtime PM by itself is not a goal. What does it change if
> it is enabled?
>
It means we could have efficient power management for this driver.
as per my understanding, it controls runtime sleep and improves power efficiency

> > I looked into imx thermal driver # drivers/thermal/imx_thermal.c
> > to enable run-time power management for exynos driver.
>
> So you have runtime PM enabled and then what happens? Where is the power
> saving? Since you did not implement the callbacks, all this should be
> explained in commit msg.
>
Ok, As per the original code, it just registers the SIMPLE_DEV_PM_OPS
with .pm = &exynos_tmu_pm
So I have made sure that suspend resume feature works correctly
 with these changes on SBC Odroid U3 and XU4.

I will try to look into setting RUNTIME_PM_OPS
or use UNIVERSAL_DEV_PM_OPS instead of SIMPLE_DEV_PM_OPS
any thought on this?

>
> Best regards,
> Krzysztof

Thanks and Regards

-Anand
Krzysztof Kozlowski May 21, 2022, 2:11 p.m. UTC | #2
On 21/05/2022 11:52, Anand Moon wrote:
> Hi Krzysztof,
> 
> On Wed, 18 May 2022 at 12:49, Krzysztof Kozlowski
> <krzysztof.kozlowski@linaro.org> wrote:
>>
>> On 17/05/2022 20:45, Anand Moon wrote:
>>> Hi Krzysztof,
>>>
>>> On Sun, 15 May 2022 at 15:18, Krzysztof Kozlowski
>>> <krzysztof.kozlowski@linaro.org> wrote:
>>>>
>>>> On 15/05/2022 08:41, Anand Moon wrote:
>>>>> Add runtime power management for exynos thermal driver.
>>>>
>>>> First of all - why? Second, I do not see it being added. Where are the
>>>> runtime callbacks?
>>>>
>>>
>>> To control runtime control PMU, did I miss something?
>>
>> Controlling runtime PM by itself is not a goal. What does it change if
>> it is enabled?
>>
> It means we could have efficient power management for this driver.
> as per my understanding, it controls runtime sleep and improves power efficiency

How? I asked - what is being changed after enabling PM - and you
answered without any specifics. Where exactly is the power saving?
Please be specific, very specific.

> 
>>> I looked into imx thermal driver # drivers/thermal/imx_thermal.c
>>> to enable run-time power management for exynos driver.
>>
>> So you have runtime PM enabled and then what happens? Where is the power
>> saving? Since you did not implement the callbacks, all this should be
>> explained in commit msg.
>>
> Ok, As per the original code, it just registers the SIMPLE_DEV_PM_OPS
> with .pm = &exynos_tmu_pm

And does nothing else, right? No benefits?

> So I have made sure that suspend resume feature works correctly
>  with these changes on SBC Odroid U3 and XU4.

How is suspend/resume related to runtime PM? Are you talking about
system suspend? What do you mean now?

> 
> I will try to look into setting RUNTIME_PM_OPS
> or use UNIVERSAL_DEV_PM_OPS instead of SIMPLE_DEV_PM_OPS
> any thought on this?

Why looking at them? You avoid giving any specific answer, so we are
repeating the same and the same. Just to be sure - maybe I don't see the
obvious stuff, so please explain also this obvious things.

Please come with specifics, because otherwise I see it as a waste of our
time.

Best regards,
Krzysztof
diff mbox series

Patch

diff --git a/drivers/thermal/samsung/exynos_tmu.c b/drivers/thermal/samsung/exynos_tmu.c
index f8a527f19383..be9b98caf2ba 100644
--- a/drivers/thermal/samsung/exynos_tmu.c
+++ b/drivers/thermal/samsung/exynos_tmu.c
@@ -20,6 +20,7 @@ 
 #include <linux/of_irq.h>
 #include <linux/platform_device.h>
 #include <linux/regulator/consumer.h>
+#include <linux/pm_runtime.h>
 
 #include <dt-bindings/thermal/thermal_exynos.h>
 
@@ -1106,6 +1107,15 @@  static int exynos_tmu_probe(struct platform_device *pdev)
 		goto err_thermal;
 	}
 
+	pm_runtime_set_active(&pdev->dev);
+	pm_runtime_set_autosuspend_delay(&pdev->dev, 1000);
+	pm_runtime_use_autosuspend(&pdev->dev);
+	pm_runtime_enable(&pdev->dev);
+
+	ret = pm_runtime_resume_and_get(&pdev->dev);
+	if (ret < 0)
+		goto disable_runtime_pm;
+
 	ret = devm_request_irq(&pdev->dev, data->irq, exynos_tmu_irq,
 		IRQF_TRIGGER_RISING | IRQF_SHARED, dev_name(&pdev->dev), data);
 	if (ret) {
@@ -1113,11 +1123,16 @@  static int exynos_tmu_probe(struct platform_device *pdev)
 		goto err_thermal;
 	}
 
+	pm_runtime_put(&pdev->dev);
+
 	exynos_tmu_control(pdev, true);
 	return 0;
 
 err_thermal:
 	thermal_zone_of_sensor_unregister(&pdev->dev, data->tzd);
+disable_runtime_pm:
+	pm_runtime_put_noidle(&pdev->dev);
+	pm_runtime_disable(&pdev->dev);
 err_clk_sec:
 	clk_disable_unprepare(data->clk_sec);
 err_sclk:
@@ -1143,6 +1158,9 @@  static int exynos_tmu_remove(struct platform_device *pdev)
 	clk_disable_unprepare(data->clk);
 	clk_disable_unprepare(data->clk_sec);
 
+	pm_runtime_put_noidle(&pdev->dev);
+	pm_runtime_disable(&pdev->dev);
+
 	if (!IS_ERR(data->regulator))
 		regulator_disable(data->regulator);
 
@@ -1151,18 +1169,25 @@  static int exynos_tmu_remove(struct platform_device *pdev)
 
 static int __maybe_unused exynos_tmu_suspend(struct device *dev)
 {
-	exynos_tmu_control(to_platform_device(dev), false);
+	struct platform_device *pdev = to_platform_device(dev);
 
-	return 0;
+	exynos_tmu_control(pdev, false);
+
+	return pm_runtime_force_suspend(&pdev->dev);
 }
 
 static int __maybe_unused exynos_tmu_resume(struct device *dev)
 {
 	struct platform_device *pdev = to_platform_device(dev);
+	int ret;
 
 	exynos_tmu_initialize(pdev);
 	exynos_tmu_control(pdev, true);
 
+	ret = pm_runtime_force_resume(&pdev->dev);
+	if (ret)
+		return ret;
+
 	return 0;
 }