Re: [PATCH v3 3/4] ASoC: mediatek: common: permit to provide dedicated regmap for irq

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Architectural Anti-Pattern: Adding a dynamically allocated per-device pointer to a global static hardware description structure.
--

commit b0a6d4d7d0d69f71dca69402decc8e6c50ee4811
Author: Christian Marangi <ansuelsmth@gmail.com>

ASoC: mediatek: common: permit to provide dedicated regmap for irq

This commit adds an extra entry in the irq_data struct to allow using a
specific regmap if defined. If it is not defined, the global AFE regmap is
used as a fallback.

> diff --git a/sound/soc/mediatek/common/mtk-base-afe.h b/sound/soc/mediatek/common/mtk-base-afe.h
> --- a/sound/soc/mediatek/common/mtk-base-afe.h
> +++ b/sound/soc/mediatek/common/mtk-base-afe.h
[ ... ]
> @@ -87,6 +87,8 @@ struct mtk_base_irq_data {
>  	int irq_clr_reg;
>  	int irq_clr_shift;
>  	int irq_status_shift;
> +
> +	struct regmap *regmap;

[Severity: High]
Does this change violate the Linux device model by adding a dynamically
allocated per-device pointer to a global static hardware description
structure?

If a devm-managed regmap is assigned to this global static array during
device probe, and the driver is unbound via sysfs, the devm framework will
free the regmap. This would leave a dangling pointer in the global irq_data
array.

Could this lead to a use-after-free if an interrupt handler directly accesses
this global array to retrieve the regmap during teardown, or if multiple
device instances exist?

Would it be safer to place the instance-specific regmap pointer in
struct mtk_base_afe_irq instead, which is properly dynamically allocated per
device instance during probe?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260528174840.28644-1-ansuelsmth@gmail.com?part=3