2021-04-08 09:11:27

by Dinghao Liu

Subject: [PATCH] media: imx: imx7-mipi-csis: Fix runtime PM imbalance in mipi_csis_s_stream

When v4l2_subdev_call() fails, a pairing PM usage counter
decrement is needed to keep the counter balanced. It's the
same for the following error paths in case 'enable' is on.

Signed-off-by: Dinghao Liu <[email protected]>
---
drivers/staging/media/imx/imx7-mipi-csis.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/staging/media/imx/imx7-mipi-csis.c b/drivers/staging/media/imx/imx7-mipi-csis.c
index a01a7364b4b9..2a3fff231a40 100644
--- a/drivers/staging/media/imx/imx7-mipi-csis.c
+++ b/drivers/staging/media/imx/imx7-mipi-csis.c
@@ -627,21 +627,26 @@ static int mipi_csis_s_stream(struct v4l2_subdev *mipi_sd, int enable)
return ret;
}
ret = v4l2_subdev_call(state->src_sd, core, s_power, 1);
- if (ret < 0)
+ if (ret < 0) {
+ pm_runtime_put_noidle(&state->pdev->dev);
return ret;
+ }
}

mutex_lock(&state->lock);
if (enable) {
if (state->flags & ST_SUSPENDED) {
ret = -EBUSY;
+ pm_runtime_put_noidle(&state->pdev->dev);
goto unlock;
}

mipi_csis_start_stream(state);
ret = v4l2_subdev_call(state->src_sd, video, s_stream, 1);
- if (ret < 0)
+ if (ret < 0) {
+ pm_runtime_put_noidle(&state->pdev->dev);
goto unlock;
+ }

mipi_csis_log_counters(state, true);

--
2.17.1

2021-04-08 13:58:23

by Rui Miguel Silva

Subject: Re: [PATCH] media: imx: imx7-mipi-csis: Fix runtime PM imbalance in mipi_csis_s_stream

Hi Liu,
Thanks for your patch.

On Thu, Apr 08, 2021 at 05:08:27PM +0800, Dinghao Liu wrote:
> When v4l2_subdev_call() fails, a pairing PM usage counter
> decrement is needed to keep the counter balanced. It's the
> same for the following error paths in case 'enable' is on.
>
> Signed-off-by: Dinghao Liu <[email protected]>
> ---
> drivers/staging/media/imx/imx7-mipi-csis.c | 9 +++++++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/staging/media/imx/imx7-mipi-csis.c b/drivers/staging/media/imx/imx7-mipi-csis.c
> index a01a7364b4b9..2a3fff231a40 100644
> --- a/drivers/staging/media/imx/imx7-mipi-csis.c
> +++ b/drivers/staging/media/imx/imx7-mipi-csis.c
> @@ -627,21 +627,26 @@ static int mipi_csis_s_stream(struct v4l2_subdev *mipi_sd, int enable)
> return ret;
> }
> ret = v4l2_subdev_call(state->src_sd, core, s_power, 1);
> - if (ret < 0)
> + if (ret < 0) {
> + pm_runtime_put_noidle(&state->pdev->dev);

I think here we should go completely pm_runtime_put to call the
mipi_csis_pm_suspend down the line, right?

> return ret;
> + }
> }
>
> mutex_lock(&state->lock);
> if (enable) {
> if (state->flags & ST_SUSPENDED) {
> ret = -EBUSY;
> + pm_runtime_put_noidle(&state->pdev->dev);

since we are in ST_SUSPENDED state, for sure the pm counter was
already 0.

> goto unlock;
> }
>
> mipi_csis_start_stream(state);
> ret = v4l2_subdev_call(state->src_sd, video, s_stream, 1);
> - if (ret < 0)
> + if (ret < 0) {
> + pm_runtime_put_noidle(&state->pdev->dev);

here also we need the pm_runtime_put, maybe just changing the unlock
tag bellow from:
if (!enable)
pm_runtime_put(&state->pdev->dev);

to
if (!enable || (ret < 0))
pm_runtime_put(&state->pdev->dev);

will not hurt the first case and will complete the suspend routine
afterward in the second case.

------
Cheers,
Rui
> goto unlock;
> + }
>
> mipi_csis_log_counters(state, true);
>
> --
> 2.17.1
>
>

2021-04-09 08:01:28

by Dinghao Liu

Subject: Re: Re: [PATCH] media: imx: imx7-mipi-csis: Fix runtime PM imbalance in mipi_csis_s_stream

> Hi Liu,
> Thanks for your patch.
>
> On Thu, Apr 08, 2021 at 05:08:27PM +0800, Dinghao Liu wrote:
> > When v4l2_subdev_call() fails, a pairing PM usage counter
> > decrement is needed to keep the counter balanced. It's the
> > same for the following error paths in case 'enable' is on.
> >
> > Signed-off-by: Dinghao Liu <[email protected]>
> > ---
> > drivers/staging/media/imx/imx7-mipi-csis.c | 9 +++++++--
> > 1 file changed, 7 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/staging/media/imx/imx7-mipi-csis.c b/drivers/staging/media/imx/imx7-mipi-csis.c
> > index a01a7364b4b9..2a3fff231a40 100644
> > --- a/drivers/staging/media/imx/imx7-mipi-csis.c
> > +++ b/drivers/staging/media/imx/imx7-mipi-csis.c
> > @@ -627,21 +627,26 @@ static int mipi_csis_s_stream(struct v4l2_subdev *mipi_sd, int enable)
> > return ret;
> > }
> > ret = v4l2_subdev_call(state->src_sd, core, s_power, 1);
> > - if (ret < 0)
> > + if (ret < 0) {
> > + pm_runtime_put_noidle(&state->pdev->dev);
>
> I think here we should go completely pm_runtime_put to call the
> mipi_csis_pm_suspend down the line, right?
>
> > return ret;
> > + }
> > }
> >
> > mutex_lock(&state->lock);
> > if (enable) {
> > if (state->flags & ST_SUSPENDED) {
> > ret = -EBUSY;
> > + pm_runtime_put_noidle(&state->pdev->dev);
>
> since we are in ST_SUSPENDED state, for sure the pm counter was
> already 0.
>
> > goto unlock;
> > }
> >
> > mipi_csis_start_stream(state);
> > ret = v4l2_subdev_call(state->src_sd, video, s_stream, 1);
> > - if (ret < 0)
> > + if (ret < 0) {
> > + pm_runtime_put_noidle(&state->pdev->dev);
>
> here also we need the pm_runtime_put, maybe just changing the unlock
> tag bellow from:
> if (!enable)
> pm_runtime_put(&state->pdev->dev);
>
> to
> if (!enable || (ret < 0))
> pm_runtime_put(&state->pdev->dev);
>
> will not hurt the first case and will complete the suspend routine
> afterward in the second case.
>

This is much clearer, thanks! I will fix this and send a new patch soon.

Regards,
Dinghao