[PATCH 0/3] media: atmel-isi: rework on the clock part and add runtime pm support

[PATCH 0/3] media: atmel-isi: rework on the clock part and add runtime pm support

[Josh Wu]

This patch series fix the peripheral clock code and enable runtime pm
support.
Aslo clean up the code which is for the compatiblity of mck.


Josh Wu (3):
  media: atmel-isi: move the peripheral clock to start/stop_stream()
    function
  media: atmel-isi: add runtime pm support
  media: atmel-isi: remove mck back compatiable code as we don't need it

 drivers/media/platform/soc_camera/atmel-isi.c | 72 +++++++++++++--------------
 1 file changed, 34 insertions(+), 38 deletions(-)

-- 
1.9.1

[PATCH 1/3] media: atmel-isi: move the peripheral clock to start/stop_stream() function

[Josh Wu]

As the clock_start/stop() use to control the mclk for the sensor not the
ISI peripheral clock.
So we move them to start/stop_stream() function.

Signed-off-by: Josh Wu <josh.wu@atmel.com>
---

 drivers/media/platform/soc_camera/atmel-isi.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/media/platform/soc_camera/atmel-isi.c b/drivers/media/platform/soc_camera/atmel-isi.c
index 1208818..eb179e7 100644
--- a/drivers/media/platform/soc_camera/atmel-isi.c
+++ b/drivers/media/platform/soc_camera/atmel-isi.c
@@ -386,6 +386,10 @@ static int start_streaming(struct vb2_queue *vq, unsigned int count)
 	struct atmel_isi *isi = ici->priv;
 	int ret;
 
+	ret = clk_prepare_enable(isi->pclk);
+	if (ret)
+		return ret;
+
 	/* Reset ISI */
 	ret = atmel_isi_wait_status(isi, WAIT_ISI_RESET);
 	if (ret < 0) {
@@ -445,6 +449,8 @@ static void stop_streaming(struct vb2_queue *vq)
 	ret = atmel_isi_wait_status(isi, WAIT_ISI_DISABLE);
 	if (ret < 0)
 		dev_err(icd->parent, "Disable ISI timed out\n");
+
+	clk_disable_unprepare(isi->pclk);
 }
 
 static struct vb2_ops isi_video_qops = {
@@ -723,14 +729,9 @@ static int isi_camera_clock_start(struct soc_camera_host *ici)
 	struct atmel_isi *isi = ici->priv;
 	int ret;
 
-	ret = clk_prepare_enable(isi->pclk);
-	if (ret)
-		return ret;
-
 	if (!IS_ERR(isi->mck)) {
 		ret = clk_prepare_enable(isi->mck);
 		if (ret) {
-			clk_disable_unprepare(isi->pclk);
 			return ret;
 		}
 	}
@@ -745,7 +746,6 @@ static void isi_camera_clock_stop(struct soc_camera_host *ici)
 
 	if (!IS_ERR(isi->mck))
 		clk_disable_unprepare(isi->mck);
-	clk_disable_unprepare(isi->pclk);
 }
 
 static unsigned int isi_camera_poll(struct file *file, poll_table *pt)
-- 
1.9.1

[PATCH 1/3] media: atmel-isi: move the peripheral clock to start/stop_stream() function

[Laurent Pinchart]

Hi Josh,

Thank you for the patch.

On Thursday 05 March 2015 13:00:59 Josh Wu wrote:
> As the clock_start/stop() use to control the mclk for the sensor not the
> ISI peripheral clock.
> So we move them to start/stop_stream() function.

Then the driver will access registers with the peripheral clock disabled, for 
instance in isi_camera_set_fmt() (calling configure_geometry), 
isi_camera_set_bus_param() or atmel_isi_probe(). Isn't that a problem ? Or are 
all registers guaranteed to be accessible (and retained) when the clock is 
disabled ?

> Signed-off-by: Josh Wu <josh.wu@atmel.com>
> ---
> 
>  drivers/media/platform/soc_camera/atmel-isi.c | 12 ++++++------
>  1 file changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/media/platform/soc_camera/atmel-isi.c
> b/drivers/media/platform/soc_camera/atmel-isi.c index 1208818..eb179e7
> 100644
> --- a/drivers/media/platform/soc_camera/atmel-isi.c
> +++ b/drivers/media/platform/soc_camera/atmel-isi.c
> @@ -386,6 +386,10 @@ static int start_streaming(struct vb2_queue *vq,
> unsigned int count) struct atmel_isi *isi = ici->priv;
>  	int ret;
> 
> +	ret = clk_prepare_enable(isi->pclk);
> +	if (ret)
> +		return ret;
> +
>  	/* Reset ISI */
>  	ret = atmel_isi_wait_status(isi, WAIT_ISI_RESET);
>  	if (ret < 0) {
> @@ -445,6 +449,8 @@ static void stop_streaming(struct vb2_queue *vq)
>  	ret = atmel_isi_wait_status(isi, WAIT_ISI_DISABLE);
>  	if (ret < 0)
>  		dev_err(icd->parent, "Disable ISI timed out\n");
> +
> +	clk_disable_unprepare(isi->pclk);
>  }
> 
>  static struct vb2_ops isi_video_qops = {
> @@ -723,14 +729,9 @@ static int isi_camera_clock_start(struct
> soc_camera_host *ici) struct atmel_isi *isi = ici->priv;
>  	int ret;
> 
> -	ret = clk_prepare_enable(isi->pclk);
> -	if (ret)
> -		return ret;
> -
>  	if (!IS_ERR(isi->mck)) {
>  		ret = clk_prepare_enable(isi->mck);
>  		if (ret) {
> -			clk_disable_unprepare(isi->pclk);
>  			return ret;
>  		}
>  	}
> @@ -745,7 +746,6 @@ static void isi_camera_clock_stop(struct soc_camera_host
> *ici)
> 
>  	if (!IS_ERR(isi->mck))
>  		clk_disable_unprepare(isi->mck);
> -	clk_disable_unprepare(isi->pclk);
>  }
> 
>  static unsigned int isi_camera_poll(struct file *file, poll_table *pt)

-- 
Regards,

Laurent Pinchart

[PATCH 1/3] media: atmel-isi: move the peripheral clock to start/stop_stream() function

[Josh Wu]

On 3/5/2015 6:39 PM, Laurent Pinchart wrote:
> Hi Josh,
>
> Thank you for the patch.
>
> On Thursday 05 March 2015 13:00:59 Josh Wu wrote:
>> As the clock_start/stop() use to control the mclk for the sensor not the
>> ISI peripheral clock.
>> So we move them to start/stop_stream() function.
> Then the driver will access registers with the peripheral clock disabled, for
> instance in isi_camera_set_fmt() (calling configure_geometry),
> isi_camera_set_bus_param() or atmel_isi_probe(). Isn't that a problem ? Or are
> all registers guaranteed to be accessible (and retained) when the clock is
> disabled ?

So far, I don't see any problem yet. But I'd like to be sure of that. 
I'll give you the feedback after more test.
Thanks for make me head up.

Best Regards,
Josh Wu

>
>> Signed-off-by: Josh Wu <josh.wu@atmel.com>
>> ---
>>
>>   drivers/media/platform/soc_camera/atmel-isi.c | 12 ++++++------
>>   1 file changed, 6 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/media/platform/soc_camera/atmel-isi.c
>> b/drivers/media/platform/soc_camera/atmel-isi.c index 1208818..eb179e7
>> 100644
>> --- a/drivers/media/platform/soc_camera/atmel-isi.c
>> +++ b/drivers/media/platform/soc_camera/atmel-isi.c
>> @@ -386,6 +386,10 @@ static int start_streaming(struct vb2_queue *vq,
>> unsigned int count) struct atmel_isi *isi = ici->priv;
>>   	int ret;
>>
>> +	ret = clk_prepare_enable(isi->pclk);
>> +	if (ret)
>> +		return ret;
>> +
>>   	/* Reset ISI */
>>   	ret = atmel_isi_wait_status(isi, WAIT_ISI_RESET);
>>   	if (ret < 0) {
>> @@ -445,6 +449,8 @@ static void stop_streaming(struct vb2_queue *vq)
>>   	ret = atmel_isi_wait_status(isi, WAIT_ISI_DISABLE);
>>   	if (ret < 0)
>>   		dev_err(icd->parent, "Disable ISI timed out\n");
>> +
>> +	clk_disable_unprepare(isi->pclk);
>>   }
>>
>>   static struct vb2_ops isi_video_qops = {
>> @@ -723,14 +729,9 @@ static int isi_camera_clock_start(struct
>> soc_camera_host *ici) struct atmel_isi *isi = ici->priv;
>>   	int ret;
>>
>> -	ret = clk_prepare_enable(isi->pclk);
>> -	if (ret)
>> -		return ret;
>> -
>>   	if (!IS_ERR(isi->mck)) {
>>   		ret = clk_prepare_enable(isi->mck);
>>   		if (ret) {
>> -			clk_disable_unprepare(isi->pclk);
>>   			return ret;
>>   		}
>>   	}
>> @@ -745,7 +746,6 @@ static void isi_camera_clock_stop(struct soc_camera_host
>> *ici)
>>
>>   	if (!IS_ERR(isi->mck))
>>   		clk_disable_unprepare(isi->mck);
>> -	clk_disable_unprepare(isi->pclk);
>>   }
>>
>>   static unsigned int isi_camera_poll(struct file *file, poll_table *pt)

[PATCH 2/3] media: atmel-isi: add runtime pm support

[Josh Wu]

The runtime pm resume/suspend will enable/disable pclk (ISI peripheral
clock).
And we need to call rumtime_pm_get/put_sync function to make pm
resume/suspend.

Signed-off-by: Josh Wu <josh.wu@atmel.com>
---

 drivers/media/platform/soc_camera/atmel-isi.c | 36 ++++++++++++++++++++++++---
 1 file changed, 32 insertions(+), 4 deletions(-)

diff --git a/drivers/media/platform/soc_camera/atmel-isi.c b/drivers/media/platform/soc_camera/atmel-isi.c
index eb179e7..4a384f1 100644
--- a/drivers/media/platform/soc_camera/atmel-isi.c
+++ b/drivers/media/platform/soc_camera/atmel-isi.c
@@ -20,6 +20,7 @@
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
 #include <linux/slab.h>
 
 #include <media/atmel-isi.h>
@@ -386,9 +387,7 @@ static int start_streaming(struct vb2_queue *vq, unsigned int count)
 	struct atmel_isi *isi = ici->priv;
 	int ret;
 
-	ret = clk_prepare_enable(isi->pclk);
-	if (ret)
-		return ret;
+	pm_runtime_get_sync(ici->v4l2_dev.dev);
 
 	/* Reset ISI */
 	ret = atmel_isi_wait_status(isi, WAIT_ISI_RESET);
@@ -450,7 +449,7 @@ static void stop_streaming(struct vb2_queue *vq)
 	if (ret < 0)
 		dev_err(icd->parent, "Disable ISI timed out\n");
 
-	clk_disable_unprepare(isi->pclk);
+	pm_runtime_put_sync(ici->v4l2_dev.dev);
 }
 
 static struct vb2_ops isi_video_qops = {
@@ -1042,6 +1041,9 @@ static int atmel_isi_probe(struct platform_device *pdev)
 	soc_host->v4l2_dev.dev	= &pdev->dev;
 	soc_host->nr		= pdev->id;
 
+	pm_suspend_ignore_children(&pdev->dev, true);
+	pm_runtime_enable(&pdev->dev);
+
 	if (isi->pdata.asd_sizes) {
 		soc_host->asd = isi->pdata.asd;
 		soc_host->asd_sizes = isi->pdata.asd_sizes;
@@ -1055,6 +1057,7 @@ static int atmel_isi_probe(struct platform_device *pdev)
 	return 0;
 
 err_register_soc_camera_host:
+	pm_runtime_disable(&pdev->dev);
 err_req_irq:
 err_ioremap:
 	vb2_dma_contig_cleanup_ctx(isi->alloc_ctx);
@@ -1067,6 +1070,30 @@ err_alloc_ctx:
 	return ret;
 }
 
+static int atmel_isi_runtime_suspend(struct device *dev)
+{
+	struct soc_camera_host *soc_host = to_soc_camera_host(dev);
+	struct atmel_isi *isi = container_of(soc_host,
+					struct atmel_isi, soc_host);
+
+	clk_disable_unprepare(isi->pclk);
+
+	return 0;
+}
+static int atmel_isi_runtime_resume(struct device *dev)
+{
+	struct soc_camera_host *soc_host = to_soc_camera_host(dev);
+	struct atmel_isi *isi = container_of(soc_host,
+					struct atmel_isi, soc_host);
+
+	return clk_prepare_enable(isi->pclk);
+}
+
+static const struct dev_pm_ops atmel_isi_dev_pm_ops = {
+	SET_RUNTIME_PM_OPS(atmel_isi_runtime_suspend,
+				atmel_isi_runtime_resume, NULL)
+};
+
 static const struct of_device_id atmel_isi_of_match[] = {
 	{ .compatible = "atmel,at91sam9g45-isi" },
 	{ }
@@ -1079,6 +1106,7 @@ static struct platform_driver atmel_isi_driver = {
 		.name = "atmel_isi",
 		.owner = THIS_MODULE,
 		.of_match_table = of_match_ptr(atmel_isi_of_match),
+		.pm	= &atmel_isi_dev_pm_ops,
 	},
 };
 
-- 
1.9.1

[PATCH 2/3] media: atmel-isi: add runtime pm support

[Laurent Pinchart]

Hi Josh,

Thank you for the patch.

On Thursday 05 March 2015 13:01:00 Josh Wu wrote:
> The runtime pm resume/suspend will enable/disable pclk (ISI peripheral
> clock).
> And we need to call rumtime_pm_get/put_sync function to make pm
> resume/suspend.
> 
> Signed-off-by: Josh Wu <josh.wu@atmel.com>
> ---
> 
>  drivers/media/platform/soc_camera/atmel-isi.c | 36 +++++++++++++++++++++---
>  1 file changed, 32 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/media/platform/soc_camera/atmel-isi.c
> b/drivers/media/platform/soc_camera/atmel-isi.c index eb179e7..4a384f1
> 100644
> --- a/drivers/media/platform/soc_camera/atmel-isi.c
> +++ b/drivers/media/platform/soc_camera/atmel-isi.c
> @@ -20,6 +20,7 @@
>  #include <linux/kernel.h>
>  #include <linux/module.h>
>  #include <linux/platform_device.h>
> +#include <linux/pm_runtime.h>
>  #include <linux/slab.h>
> 
>  #include <media/atmel-isi.h>
> @@ -386,9 +387,7 @@ static int start_streaming(struct vb2_queue *vq,
> unsigned int count) struct atmel_isi *isi = ici->priv;
>  	int ret;
> 
> -	ret = clk_prepare_enable(isi->pclk);
> -	if (ret)
> -		return ret;
> +	pm_runtime_get_sync(ici->v4l2_dev.dev);
> 
>  	/* Reset ISI */
>  	ret = atmel_isi_wait_status(isi, WAIT_ISI_RESET);
> @@ -450,7 +449,7 @@ static void stop_streaming(struct vb2_queue *vq)
>  	if (ret < 0)
>  		dev_err(icd->parent, "Disable ISI timed out\n");
> 
> -	clk_disable_unprepare(isi->pclk);
> +	pm_runtime_put_sync(ici->v4l2_dev.dev);

Is there a reason to use pm_runtime_put_sync() and not just pm_runtime_put() ?

>  }
> 
>  static struct vb2_ops isi_video_qops = {
> @@ -1042,6 +1041,9 @@ static int atmel_isi_probe(struct platform_device
> *pdev) soc_host->v4l2_dev.dev	= &pdev->dev;
>  	soc_host->nr		= pdev->id;
> 
> +	pm_suspend_ignore_children(&pdev->dev, true);
> +	pm_runtime_enable(&pdev->dev);
> +
>  	if (isi->pdata.asd_sizes) {
>  		soc_host->asd = isi->pdata.asd;
>  		soc_host->asd_sizes = isi->pdata.asd_sizes;
> @@ -1055,6 +1057,7 @@ static int atmel_isi_probe(struct platform_device
> *pdev) return 0;
> 
>  err_register_soc_camera_host:
> +	pm_runtime_disable(&pdev->dev);
>  err_req_irq:
>  err_ioremap:
>  	vb2_dma_contig_cleanup_ctx(isi->alloc_ctx);
> @@ -1067,6 +1070,30 @@ err_alloc_ctx:
>  	return ret;
>  }
> 
> +static int atmel_isi_runtime_suspend(struct device *dev)
> +{
> +	struct soc_camera_host *soc_host = to_soc_camera_host(dev);
> +	struct atmel_isi *isi = container_of(soc_host,
> +					struct atmel_isi, soc_host);
> +
> +	clk_disable_unprepare(isi->pclk);
> +
> +	return 0;
> +}
> +static int atmel_isi_runtime_resume(struct device *dev)
> +{
> +	struct soc_camera_host *soc_host = to_soc_camera_host(dev);
> +	struct atmel_isi *isi = container_of(soc_host,
> +					struct atmel_isi, soc_host);
> +
> +	return clk_prepare_enable(isi->pclk);
> +}
> +
> +static const struct dev_pm_ops atmel_isi_dev_pm_ops = {
> +	SET_RUNTIME_PM_OPS(atmel_isi_runtime_suspend,
> +				atmel_isi_runtime_resume, NULL)
> +};
> +
>  static const struct of_device_id atmel_isi_of_match[] = {
>  	{ .compatible = "atmel,at91sam9g45-isi" },
>  	{ }
> @@ -1079,6 +1106,7 @@ static struct platform_driver atmel_isi_driver = {
>  		.name = "atmel_isi",
>  		.owner = THIS_MODULE,
>  		.of_match_table = of_match_ptr(atmel_isi_of_match),
> +		.pm	= &atmel_isi_dev_pm_ops,
>  	},
>  };

-- 
Regards,

Laurent Pinchart

[PATCH 2/3] media: atmel-isi: add runtime pm support

[Josh Wu]

On 3/5/2015 6:36 PM, Laurent Pinchart wrote:
> Hi Josh,
>
> Thank you for the patch.
>
> On Thursday 05 March 2015 13:01:00 Josh Wu wrote:
>> The runtime pm resume/suspend will enable/disable pclk (ISI peripheral
>> clock).
>> And we need to call rumtime_pm_get/put_sync function to make pm
>> resume/suspend.
>>
>> Signed-off-by: Josh Wu <josh.wu@atmel.com>
>> ---
>>
>>   drivers/media/platform/soc_camera/atmel-isi.c | 36 +++++++++++++++++++++---
>>   1 file changed, 32 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/media/platform/soc_camera/atmel-isi.c
>> b/drivers/media/platform/soc_camera/atmel-isi.c index eb179e7..4a384f1
>> 100644
>> --- a/drivers/media/platform/soc_camera/atmel-isi.c
>> +++ b/drivers/media/platform/soc_camera/atmel-isi.c
>> @@ -20,6 +20,7 @@
>>   #include <linux/kernel.h>
>>   #include <linux/module.h>
>>   #include <linux/platform_device.h>
>> +#include <linux/pm_runtime.h>
>>   #include <linux/slab.h>
>>
>>   #include <media/atmel-isi.h>
>> @@ -386,9 +387,7 @@ static int start_streaming(struct vb2_queue *vq,
>> unsigned int count) struct atmel_isi *isi = ici->priv;
>>   	int ret;
>>
>> -	ret = clk_prepare_enable(isi->pclk);
>> -	if (ret)
>> -		return ret;
>> +	pm_runtime_get_sync(ici->v4l2_dev.dev);
>>
>>   	/* Reset ISI */
>>   	ret = atmel_isi_wait_status(isi, WAIT_ISI_RESET);
>> @@ -450,7 +449,7 @@ static void stop_streaming(struct vb2_queue *vq)
>>   	if (ret < 0)
>>   		dev_err(icd->parent, "Disable ISI timed out\n");
>>
>> -	clk_disable_unprepare(isi->pclk);
>> +	pm_runtime_put_sync(ici->v4l2_dev.dev);
> Is there a reason to use pm_runtime_put_sync() and not just pm_runtime_put() ?

right, for sure we can use pm_runtime_put() here.

Best Regards,
Josh Wu
>
>>   }
>>
>>   static struct vb2_ops isi_video_qops = {
>> @@ -1042,6 +1041,9 @@ static int atmel_isi_probe(struct platform_device
>> *pdev) soc_host->v4l2_dev.dev	= &pdev->dev;
>>   	soc_host->nr		= pdev->id;
>>
>> +	pm_suspend_ignore_children(&pdev->dev, true);
>> +	pm_runtime_enable(&pdev->dev);
>> +
>>   	if (isi->pdata.asd_sizes) {
>>   		soc_host->asd = isi->pdata.asd;
>>   		soc_host->asd_sizes = isi->pdata.asd_sizes;
>> @@ -1055,6 +1057,7 @@ static int atmel_isi_probe(struct platform_device
>> *pdev) return 0;
>>
>>   err_register_soc_camera_host:
>> +	pm_runtime_disable(&pdev->dev);
>>   err_req_irq:
>>   err_ioremap:
>>   	vb2_dma_contig_cleanup_ctx(isi->alloc_ctx);
>> @@ -1067,6 +1070,30 @@ err_alloc_ctx:
>>   	return ret;
>>   }
>>
>> +static int atmel_isi_runtime_suspend(struct device *dev)
>> +{
>> +	struct soc_camera_host *soc_host = to_soc_camera_host(dev);
>> +	struct atmel_isi *isi = container_of(soc_host,
>> +					struct atmel_isi, soc_host);
>> +
>> +	clk_disable_unprepare(isi->pclk);
>> +
>> +	return 0;
>> +}
>> +static int atmel_isi_runtime_resume(struct device *dev)
>> +{
>> +	struct soc_camera_host *soc_host = to_soc_camera_host(dev);
>> +	struct atmel_isi *isi = container_of(soc_host,
>> +					struct atmel_isi, soc_host);
>> +
>> +	return clk_prepare_enable(isi->pclk);
>> +}
>> +
>> +static const struct dev_pm_ops atmel_isi_dev_pm_ops = {
>> +	SET_RUNTIME_PM_OPS(atmel_isi_runtime_suspend,
>> +				atmel_isi_runtime_resume, NULL)
>> +};
>> +
>>   static const struct of_device_id atmel_isi_of_match[] = {
>>   	{ .compatible = "atmel,at91sam9g45-isi" },
>>   	{ }
>> @@ -1079,6 +1106,7 @@ static struct platform_driver atmel_isi_driver = {
>>   		.name = "atmel_isi",
>>   		.owner = THIS_MODULE,
>>   		.of_match_table = of_match_ptr(atmel_isi_of_match),
>> +		.pm	= &atmel_isi_dev_pm_ops,
>>   	},
>>   };

[PATCH 3/3] media: atmel-isi: remove mck back compatiable code as we don't need it

[Josh Wu]

The master clock should handled by sensor itself.

Signed-off-by: Josh Wu <josh.wu@atmel.com>
---

 drivers/media/platform/soc_camera/atmel-isi.c | 32 ---------------------------
 1 file changed, 32 deletions(-)

diff --git a/drivers/media/platform/soc_camera/atmel-isi.c b/drivers/media/platform/soc_camera/atmel-isi.c
index 4a384f1..50375ce 100644
--- a/drivers/media/platform/soc_camera/atmel-isi.c
+++ b/drivers/media/platform/soc_camera/atmel-isi.c
@@ -83,8 +83,6 @@ struct atmel_isi {
 	struct completion		complete;
 	/* ISI peripherial clock */
 	struct clk			*pclk;
-	/* ISI_MCK, feed to camera sensor to generate pixel clock */
-	struct clk			*mck;
 	unsigned int			irq;
 
 	struct isi_platform_data	pdata;
@@ -725,26 +723,12 @@ static void isi_camera_remove_device(struct soc_camera_device *icd)
 /* Called with .host_lock held */
 static int isi_camera_clock_start(struct soc_camera_host *ici)
 {
-	struct atmel_isi *isi = ici->priv;
-	int ret;
-
-	if (!IS_ERR(isi->mck)) {
-		ret = clk_prepare_enable(isi->mck);
-		if (ret) {
-			return ret;
-		}
-	}
-
 	return 0;
 }
 
 /* Called with .host_lock held */
 static void isi_camera_clock_stop(struct soc_camera_host *ici)
 {
-	struct atmel_isi *isi = ici->priv;
-
-	if (!IS_ERR(isi->mck))
-		clk_disable_unprepare(isi->mck);
 }
 
 static unsigned int isi_camera_poll(struct file *file, poll_table *pt)
@@ -894,7 +878,6 @@ static int atmel_isi_probe_dt(struct atmel_isi *isi,
 
 	/* Default settings for ISI */
 	isi->pdata.full_mode = 1;
-	isi->pdata.mck_hz = ISI_DEFAULT_MCLK_FREQ;
 	isi->pdata.frate = ISI_CFG1_FRATE_CAPTURE_ALL;
 
 	np = of_graph_get_next_endpoint(np, NULL);
@@ -970,21 +953,6 @@ static int atmel_isi_probe(struct platform_device *pdev)
 	INIT_LIST_HEAD(&isi->video_buffer_list);
 	INIT_LIST_HEAD(&isi->dma_desc_head);
 
-	/* ISI_MCK is the sensor master clock. It should be handled by the
-	 * sensor driver directly, as the ISI has no use for that clock. Make
-	 * the clock optional here while platforms transition to the correct
-	 * model.
-	 */
-	isi->mck = devm_clk_get(dev, "isi_mck");
-	if (!IS_ERR(isi->mck)) {
-		/* Set ISI_MCK's frequency, it should be faster than pixel
-		 * clock.
-		 */
-		ret = clk_set_rate(isi->mck, isi->pdata.mck_hz);
-		if (ret < 0)
-			return ret;
-	}
-
 	isi->p_fb_descriptors = dma_alloc_coherent(&pdev->dev,
 				sizeof(struct fbd) * MAX_BUFFER_NUM,
 				&isi->fb_descriptors_phys,
-- 
1.9.1

[PATCH 3/3] media: atmel-isi: remove mck back compatiable code as we don't need it

[Laurent Pinchart]

Hi Josh,

Thank you for the patch.

On Thursday 05 March 2015 13:01:01 Josh Wu wrote:
> The master clock should handled by sensor itself.

I like that :-)

> Signed-off-by: Josh Wu <josh.wu@atmel.com>
> ---
> 
>  drivers/media/platform/soc_camera/atmel-isi.c | 32 ------------------------
>  1 file changed, 32 deletions(-)
> 
> diff --git a/drivers/media/platform/soc_camera/atmel-isi.c
> b/drivers/media/platform/soc_camera/atmel-isi.c index 4a384f1..50375ce
> 100644
> --- a/drivers/media/platform/soc_camera/atmel-isi.c
> +++ b/drivers/media/platform/soc_camera/atmel-isi.c
> @@ -83,8 +83,6 @@ struct atmel_isi {
>  	struct completion		complete;
>  	/* ISI peripherial clock */
>  	struct clk			*pclk;
> -	/* ISI_MCK, feed to camera sensor to generate pixel clock */
> -	struct clk			*mck;
>  	unsigned int			irq;
> 
>  	struct isi_platform_data	pdata;
> @@ -725,26 +723,12 @@ static void isi_camera_remove_device(struct
> soc_camera_device *icd) /* Called with .host_lock held */
>  static int isi_camera_clock_start(struct soc_camera_host *ici)
>  {
> -	struct atmel_isi *isi = ici->priv;
> -	int ret;
> -
> -	if (!IS_ERR(isi->mck)) {
> -		ret = clk_prepare_enable(isi->mck);
> -		if (ret) {
> -			return ret;
> -		}
> -	}
> -
>  	return 0;

Would it make sense to make the clock_start and clock_stop operations optional 
in the soc-camera core ?

>  }
> 
>  /* Called with .host_lock held */
>  static void isi_camera_clock_stop(struct soc_camera_host *ici)
>  {
> -	struct atmel_isi *isi = ici->priv;
> -
> -	if (!IS_ERR(isi->mck))
> -		clk_disable_unprepare(isi->mck);
>  }
> 
>  static unsigned int isi_camera_poll(struct file *file, poll_table *pt)
> @@ -894,7 +878,6 @@ static int atmel_isi_probe_dt(struct atmel_isi *isi,
> 
>  	/* Default settings for ISI */
>  	isi->pdata.full_mode = 1;
> -	isi->pdata.mck_hz = ISI_DEFAULT_MCLK_FREQ;
>  	isi->pdata.frate = ISI_CFG1_FRATE_CAPTURE_ALL;
> 
>  	np = of_graph_get_next_endpoint(np, NULL);
> @@ -970,21 +953,6 @@ static int atmel_isi_probe(struct platform_device
> *pdev) INIT_LIST_HEAD(&isi->video_buffer_list);
>  	INIT_LIST_HEAD(&isi->dma_desc_head);
> 
> -	/* ISI_MCK is the sensor master clock. It should be handled by the
> -	 * sensor driver directly, as the ISI has no use for that clock. Make
> -	 * the clock optional here while platforms transition to the correct
> -	 * model.
> -	 */
> -	isi->mck = devm_clk_get(dev, "isi_mck");
> -	if (!IS_ERR(isi->mck)) {
> -		/* Set ISI_MCK's frequency, it should be faster than pixel
> -		 * clock.
> -		 */
> -		ret = clk_set_rate(isi->mck, isi->pdata.mck_hz);
> -		if (ret < 0)
> -			return ret;
> -	}
> -
>  	isi->p_fb_descriptors = dma_alloc_coherent(&pdev->dev,
>  				sizeof(struct fbd) * MAX_BUFFER_NUM,
>  				&isi->fb_descriptors_phys,

-- 
Regards,

Laurent Pinchart

[PATCH 3/3] media: atmel-isi: remove mck back compatiable code as we don't need it

[Josh Wu]

On 3/5/2015 6:41 PM, Laurent Pinchart wrote:
> Hi Josh,
>
> Thank you for the patch.
>
> On Thursday 05 March 2015 13:01:01 Josh Wu wrote:
>> The master clock should handled by sensor itself.
> I like that :-)
>
>> Signed-off-by: Josh Wu <josh.wu@atmel.com>
>> ---
>>
>>   drivers/media/platform/soc_camera/atmel-isi.c | 32 ------------------------
>>   1 file changed, 32 deletions(-)
>>
>> diff --git a/drivers/media/platform/soc_camera/atmel-isi.c
>> b/drivers/media/platform/soc_camera/atmel-isi.c index 4a384f1..50375ce
>> 100644
>> --- a/drivers/media/platform/soc_camera/atmel-isi.c
>> +++ b/drivers/media/platform/soc_camera/atmel-isi.c
>> @@ -83,8 +83,6 @@ struct atmel_isi {
>>   	struct completion		complete;
>>   	/* ISI peripherial clock */
>>   	struct clk			*pclk;
>> -	/* ISI_MCK, feed to camera sensor to generate pixel clock */
>> -	struct clk			*mck;
>>   	unsigned int			irq;
>>
>>   	struct isi_platform_data	pdata;
>> @@ -725,26 +723,12 @@ static void isi_camera_remove_device(struct
>> soc_camera_device *icd) /* Called with .host_lock held */
>>   static int isi_camera_clock_start(struct soc_camera_host *ici)
>>   {
>> -	struct atmel_isi *isi = ici->priv;
>> -	int ret;
>> -
>> -	if (!IS_ERR(isi->mck)) {
>> -		ret = clk_prepare_enable(isi->mck);
>> -		if (ret) {
>> -			return ret;
>> -		}
>> -	}
>> -
>>   	return 0;
> Would it make sense to make the clock_start and clock_stop operations optional
> in the soc-camera core ?
I agree. For those camera host which don't provide master clock for 
sensor, clock_start and clock_stop should be optional.

Hi, Guennadi

Do you agree with this?

Best Regards,
Josh Wu

>
>>   }
>>
>>   /* Called with .host_lock held */
>>   static void isi_camera_clock_stop(struct soc_camera_host *ici)
>>   {
>> -	struct atmel_isi *isi = ici->priv;
>> -
>> -	if (!IS_ERR(isi->mck))
>> -		clk_disable_unprepare(isi->mck);
>>   }
>>
>>   static unsigned int isi_camera_poll(struct file *file, poll_table *pt)
>> @@ -894,7 +878,6 @@ static int atmel_isi_probe_dt(struct atmel_isi *isi,
>>
>>   	/* Default settings for ISI */
>>   	isi->pdata.full_mode = 1;
>> -	isi->pdata.mck_hz = ISI_DEFAULT_MCLK_FREQ;
>>   	isi->pdata.frate = ISI_CFG1_FRATE_CAPTURE_ALL;
>>
>>   	np = of_graph_get_next_endpoint(np, NULL);
>> @@ -970,21 +953,6 @@ static int atmel_isi_probe(struct platform_device
>> *pdev) INIT_LIST_HEAD(&isi->video_buffer_list);
>>   	INIT_LIST_HEAD(&isi->dma_desc_head);
>>
>> -	/* ISI_MCK is the sensor master clock. It should be handled by the
>> -	 * sensor driver directly, as the ISI has no use for that clock. Make
>> -	 * the clock optional here while platforms transition to the correct
>> -	 * model.
>> -	 */
>> -	isi->mck = devm_clk_get(dev, "isi_mck");
>> -	if (!IS_ERR(isi->mck)) {
>> -		/* Set ISI_MCK's frequency, it should be faster than pixel
>> -		 * clock.
>> -		 */
>> -		ret = clk_set_rate(isi->mck, isi->pdata.mck_hz);
>> -		if (ret < 0)
>> -			return ret;
>> -	}
>> -
>>   	isi->p_fb_descriptors = dma_alloc_coherent(&pdev->dev,
>>   				sizeof(struct fbd) * MAX_BUFFER_NUM,
>>   				&isi->fb_descriptors_phys,

[PATCH 3/3] media: atmel-isi: remove mck back compatiable code as we don't need it

[Guennadi Liakhovetski]

Hi Josh, Laurent,

On Fri, 6 Mar 2015, Josh Wu wrote:

> On 3/5/2015 6:41 PM, Laurent Pinchart wrote:
> > Hi Josh,
> > 
> > Thank you for the patch.
> > 
> > On Thursday 05 March 2015 13:01:01 Josh Wu wrote:
> > > The master clock should handled by sensor itself.
> > I like that :-)
> > 
> > > Signed-off-by: Josh Wu <josh.wu@atmel.com>
> > > ---
> > > 
> > >   drivers/media/platform/soc_camera/atmel-isi.c | 32
> > > ------------------------
> > >   1 file changed, 32 deletions(-)
> > > 
> > > diff --git a/drivers/media/platform/soc_camera/atmel-isi.c
> > > b/drivers/media/platform/soc_camera/atmel-isi.c index 4a384f1..50375ce
> > > 100644
> > > --- a/drivers/media/platform/soc_camera/atmel-isi.c
> > > +++ b/drivers/media/platform/soc_camera/atmel-isi.c
> > > @@ -83,8 +83,6 @@ struct atmel_isi {
> > >   	struct completion		complete;
> > >   	/* ISI peripherial clock */
> > >   	struct clk			*pclk;
> > > -	/* ISI_MCK, feed to camera sensor to generate pixel clock */
> > > -	struct clk			*mck;
> > >   	unsigned int			irq;
> > > 
> > >   	struct isi_platform_data	pdata;
> > > @@ -725,26 +723,12 @@ static void isi_camera_remove_device(struct
> > > soc_camera_device *icd) /* Called with .host_lock held */
> > >   static int isi_camera_clock_start(struct soc_camera_host *ici)
> > >   {
> > > -	struct atmel_isi *isi = ici->priv;
> > > -	int ret;
> > > -
> > > -	if (!IS_ERR(isi->mck)) {
> > > -		ret = clk_prepare_enable(isi->mck);
> > > -		if (ret) {
> > > -			return ret;
> > > -		}
> > > -	}
> > > -
> > >   	return 0;
> > Would it make sense to make the clock_start and clock_stop operations
> > optional
> > in the soc-camera core ?
> I agree. For those camera host which don't provide master clock for sensor,
> clock_start and clock_stop should be optional.
> 
> Hi, Guennadi
> 
> Do you agree with this?

Yes, sure, we can do this. Would anyone like to prepare a patch?

Thanks
Guennadi

[PATCH 3/3] media: atmel-isi: remove mck back compatiable code as we don't need it

[Laurent Pinchart]

On Friday 06 March 2015 21:25:36 Guennadi Liakhovetski wrote:
> On Fri, 6 Mar 2015, Josh Wu wrote:
> > On 3/5/2015 6:41 PM, Laurent Pinchart wrote:
> >> On Thursday 05 March 2015 13:01:01 Josh Wu wrote:
> >>> The master clock should handled by sensor itself.
> >> 
> >> I like that :-)
> >> 
> >>> Signed-off-by: Josh Wu <josh.wu@atmel.com>
> >>> ---
> >>> 
> >>>   drivers/media/platform/soc_camera/atmel-isi.c | 32 -------------------
> >>>   1 file changed, 32 deletions(-)
> >>> 
> >>> diff --git a/drivers/media/platform/soc_camera/atmel-isi.c
> >>> b/drivers/media/platform/soc_camera/atmel-isi.c index 4a384f1..50375ce
> >>> 100644
> >>> --- a/drivers/media/platform/soc_camera/atmel-isi.c
> >>> +++ b/drivers/media/platform/soc_camera/atmel-isi.c
> >>> @@ -83,8 +83,6 @@ struct atmel_isi {
> >>> 
> >>>   	struct completion		complete;
> >>>   	/* ISI peripherial clock */
> >>>   	struct clk			*pclk;
> >>> 
> >>> -	/* ISI_MCK, feed to camera sensor to generate pixel clock */
> >>> -	struct clk			*mck;
> >>> 
> >>>   	unsigned int			irq;
> >>>   	
> >>>   	struct isi_platform_data	pdata;
> >>> 
> >>> @@ -725,26 +723,12 @@ static void isi_camera_remove_device(struct
> >>> soc_camera_device *icd) /* Called with .host_lock held */
> >>> 
> >>>   static int isi_camera_clock_start(struct soc_camera_host *ici)
> >>>   {
> >>> 
> >>> -	struct atmel_isi *isi = ici->priv;
> >>> -	int ret;
> >>> -
> >>> -	if (!IS_ERR(isi->mck)) {
> >>> -		ret = clk_prepare_enable(isi->mck);
> >>> -		if (ret) {
> >>> -			return ret;
> >>> -		}
> >>> -	}
> >>> -
> >>> 
> >>>   	return 0;
> >> 
> >> Would it make sense to make the clock_start and clock_stop operations
> >> optional in the soc-camera core ?
> > 
> > I agree. For those camera host which don't provide master clock for
> > sensor, clock_start and clock_stop should be optional.
> > 
> > Hi, Guennadi
> > 
> > Do you agree with this?
> 
> Yes, sure, we can do this. Would anyone like to prepare a patch?

Josh, would you like to do that, or should I give it a go ?

-- 
Regards,

Laurent Pinchart

[PATCH 3/3] media: atmel-isi: remove mck back compatiable code as we don't need it

[Josh Wu]

Hi, Laurent

On 3/8/2015 8:28 AM, Laurent Pinchart wrote:
> On Friday 06 March 2015 21:25:36 Guennadi Liakhovetski wrote:
>> On Fri, 6 Mar 2015, Josh Wu wrote:
>>> On 3/5/2015 6:41 PM, Laurent Pinchart wrote:
>>>> On Thursday 05 March 2015 13:01:01 Josh Wu wrote:
>>>>> The master clock should handled by sensor itself.
>>>> I like that :-)
>>>>
>>>>> Signed-off-by: Josh Wu <josh.wu@atmel.com>
>>>>> ---
>>>>>
>>>>>    drivers/media/platform/soc_camera/atmel-isi.c | 32 -------------------
>>>>>    1 file changed, 32 deletions(-)
>>>>>
>>>>> diff --git a/drivers/media/platform/soc_camera/atmel-isi.c
>>>>> b/drivers/media/platform/soc_camera/atmel-isi.c index 4a384f1..50375ce
>>>>> 100644
>>>>> --- a/drivers/media/platform/soc_camera/atmel-isi.c
>>>>> +++ b/drivers/media/platform/soc_camera/atmel-isi.c
>>>>> @@ -83,8 +83,6 @@ struct atmel_isi {
>>>>>
>>>>>    	struct completion		complete;
>>>>>    	/* ISI peripherial clock */
>>>>>    	struct clk			*pclk;
>>>>>
>>>>> -	/* ISI_MCK, feed to camera sensor to generate pixel clock */
>>>>> -	struct clk			*mck;
>>>>>
>>>>>    	unsigned int			irq;
>>>>>    	
>>>>>    	struct isi_platform_data	pdata;
>>>>>
>>>>> @@ -725,26 +723,12 @@ static void isi_camera_remove_device(struct
>>>>> soc_camera_device *icd) /* Called with .host_lock held */
>>>>>
>>>>>    static int isi_camera_clock_start(struct soc_camera_host *ici)
>>>>>    {
>>>>>
>>>>> -	struct atmel_isi *isi = ici->priv;
>>>>> -	int ret;
>>>>> -
>>>>> -	if (!IS_ERR(isi->mck)) {
>>>>> -		ret = clk_prepare_enable(isi->mck);
>>>>> -		if (ret) {
>>>>> -			return ret;
>>>>> -		}
>>>>> -	}
>>>>> -
>>>>>
>>>>>    	return 0;
>>>> Would it make sense to make the clock_start and clock_stop operations
>>>> optional in the soc-camera core ?
>>> I agree. For those camera host which don't provide master clock for
>>> sensor, clock_start and clock_stop should be optional.
>>>
>>> Hi, Guennadi
>>>
>>> Do you agree with this?
>> Yes, sure, we can do this. Would anyone like to prepare a patch?
> Josh, would you like to do that, or should I give it a go ?
>

Yes, you can do that if you have time. ;-)

Best Regards,
Josh Wu