Re: [PATCH v5] soc: qcom: apr: Add check for idr_alloc and of_property_read_string_index

[Bjorn Andersson]

On Fri, Oct 07, 2022 at 11:32:40AM +0800, Jiasheng Jiang wrote:
> As idr_alloc() and of_property_read_string_index() can return negative
> numbers, it should be better to check the return value and deal with
> the exception.
> Therefore, it should be better to use goto statement to stop and return
> error.
> 
> Fixes: 6adba21eb434 ("soc: qcom: Add APR bus driver")
> Signed-off-by: Jiasheng Jiang <jiasheng@xxxxxxxxxxx>
> ---
> Changelog:
> 
> v4 -> v5:
> 
> 1. Change dev_err and goto statements.
> 
> v3 -> v4:
> 
> 1. Change the title and remove the kfree.
> 
> v2 -> v3:
> 
> 1. Change the title and use goto statement to deal with the exception.
> 
> v1 -> v2:
> 
> 1. Add dev_err and put_device in order to maintain the code consistency.
> ---
>  drivers/soc/qcom/apr.c | 17 ++++++++++++++---
>  1 file changed, 14 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/soc/qcom/apr.c b/drivers/soc/qcom/apr.c
> index b4046f393575..a942c8a8dde2 100644
> --- a/drivers/soc/qcom/apr.c
> +++ b/drivers/soc/qcom/apr.c
> @@ -454,11 +454,21 @@ static int apr_add_device(struct device *dev, struct device_node *np,
>  	adev->dev.driver = NULL;
>  
>  	spin_lock(&apr->svcs_lock);
> -	idr_alloc(&apr->svcs_idr, svc, svc_id, svc_id + 1, GFP_ATOMIC);
> +	ret = idr_alloc(&apr->svcs_idr, svc, svc_id, svc_id + 1, GFP_ATOMIC);
>  	spin_unlock(&apr->svcs_lock);
> +	if (ret < 0) {
> +		dev_err(dev, "idr_alloc failed: %d\n", ret);
> +		put_device(&adev->dev);
> +		goto out;
> +	}
>  
> -	of_property_read_string_index(np, "qcom,protection-domain",
> -				      1, &adev->service_path);
> +	ret = of_property_read_string_index(np, "qcom,protection-domain",
> +					    1, &adev->service_path);
> +	if (ret < 0) {
> +		dev_err(dev, "Failed to read second value of qcom,protection-domain\n");
> +		put_device(&adev->dev);
> +		goto out;
> +	}
>  
>  	dev_info(dev, "Adding APR/GPR dev: %s\n", dev_name(&adev->dev));
>  
> @@ -468,6 +478,7 @@ static int apr_add_device(struct device *dev, struct device_node *np,
>  		put_device(&adev->dev);

This put_device() rolls back the call to device_register(). It should
therefor not be replicated in the newly introduced error paths.

Other than that, this looks good to me.

Thanks,
Bjorn

>  	}
>  
> +out:
>  	return ret;
>  }
>  
> -- 
> 2.25.1
>