[libcamera-devel] [PATCH] libcamera: pipeline: ipu3: Stop IPA before stopping devices

[Niklas Söderlund]

Hi Kieran,

On 2021-03-08 16:29:15 +0000, Kieran Bingham wrote:
> The IPA should be stopped before the hardware devices to ensure that
> all asynchronous actions have completed within the IPA before resources
> are removed and released.
> 
> Signed-off-by: Kieran Bingham <kieran.bingham at ideasonboard.com>
> ---
>  src/libcamera/pipeline/ipu3/ipu3.cpp | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/src/libcamera/pipeline/ipu3/ipu3.cpp b/src/libcamera/pipeline/ipu3/ipu3.cpp
> index 2b4d31500533..498f85634a83 100644
> --- a/src/libcamera/pipeline/ipu3/ipu3.cpp
> +++ b/src/libcamera/pipeline/ipu3/ipu3.cpp
> @@ -764,13 +764,13 @@ void PipelineHandlerIPU3::stop(Camera *camera)
>  	IPU3CameraData *data = cameraData(camera);
>  	int ret = 0;
>  
> +	data->ipa_->stop();
> +
>  	ret |= data->imgu_->stop();
>  	ret |= data->cio2_.stop();
>  	if (ret)
>  		LOG(IPU3, Warning) << "Failed to stop camera " << camera->id();
>  
> -	data->ipa_->stop();
> -

IIRC correctly this order was picked because before the cio2 is stopped 
buffers may still be dequeued and therefor handed to the IPA for 
processing. This was however done before the new IPC change and with a 
mock IPA. How does this work with the new IPC/IPA if the pipeline 
handler tries to post buffers to be processed before the CIO2 is 
stopped?


>  	freeBuffers(camera);

Should not all resources shared between pipeline and IPA be available 
until the buffers are freed here?

>  }
>  
> -- 
> 2.25.1
> 
> _______________________________________________
> libcamera-devel mailing list
> libcamera-devel at lists.libcamera.org
> https://lists.libcamera.org/listinfo/libcamera-devel

-- 
Regards,
Niklas Söderlund