Re: [RFC PATCH 4/4] interconnect: qcom: sdm845: Split qnodes into their respective NoCs

[Georgi Djakov]

Hi David,

On 17.10.19 Ð. 5:20 Ñ., David Dai wrote:
> In order to better represent the hardware and its different Network-On-Chip
> devices, split the sdm845 provider driver into NoC specific providers.
> Remove duplicate functionality already provided by the icc rpmh and
> bcm voter drivers to calculate and commit bandwidth requests to hardware.
> 
> Signed-off-by: David Dai <daidavid1@xxxxxxxxxxxxxx>
> ---
>  drivers/interconnect/qcom/sdm845.c | 727 +++++++++++--------------------------
>  1 file changed, 206 insertions(+), 521 deletions(-)
> 
> diff --git a/drivers/interconnect/qcom/sdm845.c b/drivers/interconnect/qcom/sdm845.c
> index 502a6c2..a731f4d 100644
> --- a/drivers/interconnect/qcom/sdm845.c
> +++ b/drivers/interconnect/qcom/sdm845.c
[..]
>  static int qnoc_probe(struct platform_device *pdev)
>  {
> @@ -808,6 +480,12 @@ static int qnoc_probe(struct platform_device *pdev)
>  	qp->bcms = desc->bcms;
>  	qp->num_bcms = desc->num_bcms;
>  
> +	qp->voter = of_bcm_voter_get(qp->dev, NULL);

I assume that we could have a second optional bcm-voter? The
"qcom,bcm-voter-names" DT property is not used anywhere, is it needed? Maybe
give an example in patch 1.

> +	if (IS_ERR(qp->voter)) {
> +		dev_err(&pdev->dev, "bcm_voter err:%d\n", PTR_ERR(qp->voter));

Should be %ld

> +		return PTR_ERR(qp->voter);
> +	}
> +
>  	ret = icc_provider_add(provider);
>  	if (ret) {
>  		dev_err(&pdev->dev, "error adding interconnect provider\n");

Nit: I would also put patch 2/4 at the end of the series.

Thanks,
Georgi