[PATCH] media: ov7740: fix runtime pm initialization

[PATCH] media: ov7740: fix runtime pm initialization

[Akinobu Mita]

The runtime PM of this device is enabled after v4l2_ctrl_handler_setup(),
and this makes this device's runtime PM usage count a negative value.

The ov7740_set_ctrl() tries to do something only if the device's runtime
PM usage counter is nonzero.

ov7740_set_ctrl()
{
	if (!pm_runtime_get_if_in_use(&client->dev))
		return 0;

	<do something>;

	pm_runtime_put(&client->dev);

	return ret;
}

However, the ov7740_set_ctrl() is called by v4l2_ctrl_handler_setup()
while the runtime PM of this device is not yet enabled.  In this case,
the pm_runtime_get_if_in_use() returns -EINVAL (!= 0).

Therefore we can't bail out of this function and the usage count is
decreased by pm_runtime_put() without increment.

This fixes this problem by enabling the runtime PM of this device before
v4l2_ctrl_handler_setup() so that the ov7740_set_ctrl() is always called
when the runtime PM is enabled.

Cc: Eugen Hristev <eugen.hristev@microchip.com>
Cc: Wenyou Yang <wenyou.yang@microchip.com>
Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
---
I don't have the ov7740 device, but I saw the same problem with the
mt9m001 device when I was adding the runtime PM support for it.

Eugen Hristev reported the problem with ov7740.

https://www.mail-archive.com/linux-media@vger.kernel.org/msg144540.html

I suspect that it is related to this runtime PM problem.

There seems to be the same problem in other devices.  I would like to see
if this patch actually fix the ov7740's problem and then propagate other
devices.

 drivers/media/i2c/ov7740.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/media/i2c/ov7740.c b/drivers/media/i2c/ov7740.c
index 177688a..8835b83 100644
--- a/drivers/media/i2c/ov7740.c
+++ b/drivers/media/i2c/ov7740.c
@@ -1101,6 +1101,9 @@ static int ov7740_probe(struct i2c_client *client,
 	if (ret)
 		return ret;
 
+	pm_runtime_set_active(&client->dev);
+	pm_runtime_enable(&client->dev);
+
 	ret = ov7740_detect(ov7740);
 	if (ret)
 		goto error_detect;
@@ -1123,8 +1126,6 @@ static int ov7740_probe(struct i2c_client *client,
 	if (ret)
 		goto error_async_register;
 
-	pm_runtime_set_active(&client->dev);
-	pm_runtime_enable(&client->dev);
 	pm_runtime_idle(&client->dev);
 
 	return 0;
@@ -1134,6 +1135,8 @@ static int ov7740_probe(struct i2c_client *client,
 error_init_controls:
 	ov7740_free_controls(ov7740);
 error_detect:
+	pm_runtime_disable(&client->dev);
+	pm_runtime_set_suspended(&client->dev);
 	ov7740_set_power(ov7740, 0);
 	media_entity_cleanup(&ov7740->subdev.entity);
 
-- 
2.7.4

Re: [PATCH] media: ov7740: fix runtime pm initialization

[Eugen.Hristev]

On 17.02.2019 17:17, Akinobu Mita wrote:
> The runtime PM of this device is enabled after v4l2_ctrl_handler_setup(),
> and this makes this device's runtime PM usage count a negative value.
> 
> The ov7740_set_ctrl() tries to do something only if the device's runtime
> PM usage counter is nonzero.
> 
> ov7740_set_ctrl()
> {
> 	if (!pm_runtime_get_if_in_use(&client->dev))
> 		return 0;
> 
> 	<do something>;
> 
> 	pm_runtime_put(&client->dev);
> 
> 	return ret;
> }
> 
> However, the ov7740_set_ctrl() is called by v4l2_ctrl_handler_setup()
> while the runtime PM of this device is not yet enabled.  In this case,
> the pm_runtime_get_if_in_use() returns -EINVAL (!= 0).
> 
> Therefore we can't bail out of this function and the usage count is
> decreased by pm_runtime_put() without increment.
> 
> This fixes this problem by enabling the runtime PM of this device before
> v4l2_ctrl_handler_setup() so that the ov7740_set_ctrl() is always called
> when the runtime PM is enabled.
> 
> Cc: Eugen Hristev <eugen.hristev@microchip.com>
> Cc: Wenyou Yang <wenyou.yang@microchip.com>
> Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
> Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
> ---
> I don't have the ov7740 device, but I saw the same problem with the
> mt9m001 device when I was adding the runtime PM support for it.
> 
> Eugen Hristev reported the problem with ov7740.

Hi,

This fixes my issue with this sensor in latest mediatree
For anything's worth, for Atmel ISC + ov7740, with my patch,

Tested-by: Eugen Hristev <eugen.hristev@microchip.com>

Thanks

> 
> https://www.mail-archive.com/linux-media@vger.kernel.org/msg144540.html
> 
> I suspect that it is related to this runtime PM problem.
> 
> There seems to be the same problem in other devices.  I would like to see
> if this patch actually fix the ov7740's problem and then propagate other
> devices.
> 
>   drivers/media/i2c/ov7740.c | 7 +++++--
>   1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/media/i2c/ov7740.c b/drivers/media/i2c/ov7740.c
> index 177688a..8835b83 100644
> --- a/drivers/media/i2c/ov7740.c
> +++ b/drivers/media/i2c/ov7740.c
> @@ -1101,6 +1101,9 @@ static int ov7740_probe(struct i2c_client *client,
>   	if (ret)
>   		return ret;
>   
> +	pm_runtime_set_active(&client->dev);
> +	pm_runtime_enable(&client->dev);
> +
>   	ret = ov7740_detect(ov7740);
>   	if (ret)
>   		goto error_detect;
> @@ -1123,8 +1126,6 @@ static int ov7740_probe(struct i2c_client *client,
>   	if (ret)
>   		goto error_async_register;
>   
> -	pm_runtime_set_active(&client->dev);
> -	pm_runtime_enable(&client->dev);
>   	pm_runtime_idle(&client->dev);
>   
>   	return 0;
> @@ -1134,6 +1135,8 @@ static int ov7740_probe(struct i2c_client *client,
>   error_init_controls:
>   	ov7740_free_controls(ov7740);
>   error_detect:
> +	pm_runtime_disable(&client->dev);
> +	pm_runtime_set_suspended(&client->dev);
>   	ov7740_set_power(ov7740, 0);
>   	media_entity_cleanup(&ov7740->subdev.entity);
>   
>