Re: [PATCH v1 4/6] mfd: lm3533: set DMA mask

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Bypassing the kernel DMA API by directly modifying 'dev.dma_mask' and 'dev.coherent_dma_mask' to fake a 0-bit DMA mask.
--

commit e60afe532ed2b9233017faa2b381f88d84566aea
Author: Svyatoslav Ryhel <clamor95@gmail.com>

mfd: lm3533: set DMA mask

This commit explicitly sets the dma_mask and coherent_dma_mask of the i2c
client device to zero. This is done to suppress a "DMA mask not set"
platform warning, as the device and its children do not use DMA.

> diff --git a/drivers/mfd/lm3533-core.c b/drivers/mfd/lm3533-core.c
> index 131eb1a1c8eba..c7914afd564cf 100644
> --- a/drivers/mfd/lm3533-core.c
> +++ b/drivers/mfd/lm3533-core.c
[ ... ]
> @@ -499,6 +499,10 @@ static int lm3533_i2c_probe(struct i2c_client *i2c)
>  			lm3533->have_leds = true;
>  	}
>  
> +	/* Parent I2C controller uses DMA, LM3533 and child devices do not */
> +	i2c->dev.coherent_dma_mask = 0;
> +	i2c->dev.dma_mask = &i2c->dev.coherent_dma_mask;

Are we bypassing the kernel DMA API here by manually modifying the struct
device fields instead of using a helper like dma_coerce_mask_and_coherent()?

Directly assigning to dev.dma_mask and dev.coherent_dma_mask circumvents
architecture-specific DMA validation.

Additionally, is a 0-bit DMA mask conceptually valid? Standard validation
functions might reject a zero mask. Bypassing the API to inject an invalid
mask solely to silence a subsystem warning could misrepresent hardware
capabilities and violate architectural layering.

> +
>  	return lm3533_device_init(lm3533);
>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260517074306.30937-1-clamor95@gmail.com?part=4