[libcamera-devel] [PATCH 3/5] libcamera: ipu3-cio2: Discover VCMs through ancillary links

[Laurent Pinchart]

Hi Daniel,

Thank you for the patch.

On Fri, Nov 26, 2021 at 12:31:16AM +0000, Daniel Scally wrote:
> Rather than attempting to discover VCMs via model name matching we
> should follow the ancillary links that define the relationship
> between the two devices to locate any entities linked to the sensor.
> Where we have linked entities with the function MEDIA_ENT_F_LENS we
> can then create an instance of CameraLens to represent it.
> 
> Signed-off-by: Daniel Scally <djrscally at gmail.com>

Given that Han-lin's patch that add model-based matching hasn't been
merged yet, I think this should be rebased on top of the master branch
directly. There's no need to merge model-based matching to replace it
immediately after.

> ---
> 
> Alternatively rather than replace the matching on model names we could try
> the ancillary links first and then simply guard the existing model matching
> with an if (!lens_) ...

I think ancillary links are good enough, there's no need for model-based
matching.

>  src/libcamera/pipeline/ipu3/cio2.cpp | 45 +++++++++++++++-------------
>  1 file changed, 25 insertions(+), 20 deletions(-)
> 
> diff --git a/src/libcamera/pipeline/ipu3/cio2.cpp b/src/libcamera/pipeline/ipu3/cio2.cpp
> index 59b2f586..169e7b54 100644
> --- a/src/libcamera/pipeline/ipu3/cio2.cpp
> +++ b/src/libcamera/pipeline/ipu3/cio2.cpp
> @@ -161,30 +161,35 @@ int CIO2Device::init(const MediaDevice *media, unsigned int index)
>  	}
>  
>  	/*
> -	 * \todo Read the lens model from the sensor itself or from a device
> -	 * database. For now use default values taken from ChromeOS database.
> +	 * Sensors sometimes have ancillary devices such as a Lens or Flash
> +	 * that could be linked to the MediaEntity - search for and handle
> +	 * any such device (for now, we can only handle MEDIA_ENT_F_LENS)
> +	 *
> +	 * \todo Handle MEDIA_ENT_F_FLASH too.
>  	 */
> -	static std::unordered_map<std::string, std::string> sensorLens = {
> -		{ "ov13858", "dw9714" },
> -		{ "imx258", "dw9807" },
> -		{ "imx355", "ak7375" }
> -	};
> -
> -	auto it = sensorLens.find(sensor_->model());
> -	if (it != sensorLens.end()) {
> -		const std::vector<MediaEntity *> &entities = media->entities();
> -		for (auto ent : entities) {
> -			if (ent->function() == MEDIA_ENT_F_LENS) {
> -				lens_ = std::make_unique<CameraLens>(ent);
> +	const std::vector<MediaLink *> &ancillary_links = sensorEntity->ancillary_links();
> +	if (!ancillary_links.empty()) {
> +		for (auto it : ancillary_links) {
> +			MediaEntity *ancillaryEntity = it->ancillary();
> +
> +			switch (ancillaryEntity->function()) {
> +			case MEDIA_ENT_F_LENS:
> +				lens_ = std::make_unique<CameraLens>(ancillaryEntity);
>  				ret = lens_->init();
> -				if (!ret && lens_->model() == it->second) {
> -					break;
> +				if (ret) {
> +					LOG(IPU3, Error)
> +						<< "Error during CameraLens init";
> +					return ret;
>  				}
> -				lens_.reset();
> +				break;
> +			case MEDIA_ENT_F_FLASH:
> +				LOG(IPU3, Warning)
> +					<< "Flash not yet supported";
> +				break;
> +			default:
> +				LOG(IPU3, Warning)
> +					<< "Unsupported entity function";
>  			}

Could we simplify the pipeline handlers a bit by moving at least the
lookup of the VCM entity to the CameraSensor class ?

> -			if (!lens_)
> -				LOG(IPU3, Warning) << "Lens device "
> -						   << it->second << " not found";
>  		}
>  	}
>  

-- 
Regards,

Laurent Pinchart