Re: [PATCH] dmaengine: stm32-mdma: Add missing check for device_property_read_u32_array

[Amelie Delaunay]

On 1/28/26 05:23, Chen Ni wrote:
> Add check for the return value of device_property_read_u32_array() and
> return the error if it fails in order to catch the error.
>
> Signed-off-by: Chen Ni <nichen@xxxxxxxxxxx>
> ---
>   drivers/dma/stm32/stm32-mdma.c | 8 +++++---
>   1 file changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/dma/stm32/stm32-mdma.c b/drivers/dma/stm32/stm32-mdma.c
> index b87d41b234df..f833724cf42c 100644
> --- a/drivers/dma/stm32/stm32-mdma.c
> +++ b/drivers/dma/stm32/stm32-mdma.c
> @@ -1630,9 +1630,11 @@ static int stm32_mdma_probe(struct platform_device *pdev)
>   
>   	dmadev->nr_channels = nr_channels;
>   	dmadev->nr_requests = nr_requests;
> -	device_property_read_u32_array(&pdev->dev, "st,ahb-addr-masks",
> -				       dmadev->ahb_addr_masks,
> -				       count);
> +	ret = device_property_read_u32_array(&pdev->dev, "st,ahb-addr-masks",
> +					     dmadev->ahb_addr_masks,
> +					     count);
> +	if (ret)
> +		return ret;

Your patch breaks MDMA on STM32MP1x platforms because 
"st,ahb-addr-masks" is an optional property. That is why the return 
value of device_property_read_u32_array() is intentionally ignored.

Count is used to check the presence of the property and to allocate 
dmadev->ahb_addr_masks accordingly:
	count = device_property_count_u32(&pdev->dev, "st,ahb-addr-masks");
	if (count < 0)
		count = 0;
	[...]
	dmadev->nr_ahb_addr_masks = count;

If the property is not present, dmadev->nr_ahb_addr_masks is set to 0, 
and MDMA behaves correctly.

With your patch, when the property is missing, the MDMA probe fails 
because device_property_read_u32_array() returns -EINVAL. This is 
incorrect since "st,ahb-addr-masks" is an optional property, not a 
required one.

Regards,
Amelie

>   
>   	dmadev->base = devm_platform_ioremap_resource(pdev, 0);
>   	if (IS_ERR(dmadev->base))