Re: [RESEND 2 PATCH] media: marvell-ccic: Add support for runtime PM

[Sakari Ailus]

Hi Lubomir,

Thanks for the patch.

On Mon, Jun 01, 2020 at 09:21:24PM +0200, Lubomir Rintel wrote:
> On MMP3, the camera block lives on na separate power island. We want to

s/on \Kn//

> turn it off if the CCIC is not in use to conserve power.
> 
> Signed-off-by: Lubomir Rintel <lkundrak@xxxxx>
> ---
>  drivers/media/platform/marvell-ccic/mcam-core.c  |  3 +++
>  drivers/media/platform/marvell-ccic/mmp-driver.c | 12 ++++++++----
>  2 files changed, 11 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/media/platform/marvell-ccic/mcam-core.c b/drivers/media/platform/marvell-ccic/mcam-core.c
> index 09775b6624c6b..c2cd1d461bd06 100644
> --- a/drivers/media/platform/marvell-ccic/mcam-core.c
> +++ b/drivers/media/platform/marvell-ccic/mcam-core.c
> @@ -24,6 +24,7 @@
>  #include <linux/clk.h>
>  #include <linux/clk-provider.h>
>  #include <linux/videodev2.h>
> +#include <linux/pm_runtime.h>
>  #include <media/v4l2-device.h>
>  #include <media/v4l2-ioctl.h>
>  #include <media/v4l2-ctrls.h>
> @@ -901,6 +902,7 @@ static void mcam_clk_enable(struct mcam_camera *mcam)

The clocks look like something that rather should be controlled by the
device's runtime PM state. Could it be put behind suspend/resume callbacks?

>  {
>  	unsigned int i;
>  
> +	pm_runtime_get_sync(mcam->dev);
>  	for (i = 0; i < NR_MCAM_CLK; i++) {
>  		if (!IS_ERR(mcam->clk[i]))
>  			clk_prepare_enable(mcam->clk[i]);
> @@ -915,6 +917,7 @@ static void mcam_clk_disable(struct mcam_camera *mcam)
>  		if (!IS_ERR(mcam->clk[i]))
>  			clk_disable_unprepare(mcam->clk[i]);
>  	}
> +	pm_runtime_put(mcam->dev);
>  }
>  
>  /* ---------------------------------------------------------------------- */
> diff --git a/drivers/media/platform/marvell-ccic/mmp-driver.c b/drivers/media/platform/marvell-ccic/mmp-driver.c
> index 92b92255dac66..eec482d16805b 100644
> --- a/drivers/media/platform/marvell-ccic/mmp-driver.c
> +++ b/drivers/media/platform/marvell-ccic/mmp-driver.c
> @@ -24,6 +24,7 @@
>  #include <linux/list.h>
>  #include <linux/pm.h>
>  #include <linux/clk.h>
> +#include <linux/pm_runtime.h>
>  
>  #include "mcam-core.h"
>  
> @@ -313,10 +314,12 @@ static int mmpcam_probe(struct platform_device *pdev)
>  	cam->irq = res->start;
>  	ret = devm_request_irq(&pdev->dev, cam->irq, mmpcam_irq, IRQF_SHARED,
>  					"mmp-camera", mcam);
> -	if (ret == 0) {
> -		mmpcam_add_device(cam);
> -		return 0;
> -	}
> +	if (ret)
> +		goto out;
> +
> +	mmpcam_add_device(cam);
> +	pm_runtime_enable(&pdev->dev);

If you enable runtime PM here, there's a chance the clocks have been
prepared and enabled before this.

How about moving registerin clocks after this instead? Looking at the
current code, they are registered too early.

> +	return 0;
>  
>  out:
>  	fwnode_handle_put(mcam->asd.match.fwnode);
> @@ -332,6 +335,7 @@ static int mmpcam_remove(struct mmp_camera *cam)
>  
>  	mmpcam_remove_device(cam);
>  	mccic_shutdown(mcam);
> +	pm_runtime_force_suspend(mcam->dev);
>  	return 0;
>  }
>  

-- 
Kind regards,

Sakari Ailus