Re: [PATCH v6 05/11] mfd: sec: set DMA coherent mask

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Setting the parent device's `dma_mask` pointer to its own `coherent_dma_mask` introduces a shared-state data race across MFD sub-devices and circumvents DMA API validation.
--

commit ed135db5f1241e8cf0252f12ff10c2221996d7ef
Author: Kaustabh Chakraborty <kauschluss@disroot.org>

mfd: sec: set DMA coherent mask

This commit manually sets the coherent_dma_mask to 0 and assigns the
dma_mask pointer to it for the SEC PMIC device. This is done to silence
warnings about missing DMA masks on sub-devices since the device does not
use DMA.

> diff --git a/drivers/mfd/sec-common.c b/drivers/mfd/sec-common.c
> index ad1c7de83a620..afc9e691bb67c 100644
> --- a/drivers/mfd/sec-common.c
> +++ b/drivers/mfd/sec-common.c
> @@ -215,6 +215,9 @@ int sec_pmic_probe(struct device *dev, int device_type, unsigned int irq,
>  	if (IS_ERR(irq_data))
>  		return PTR_ERR(irq_data);
>  
> +	dev->coherent_dma_mask = 0;
> +	dev->dma_mask = &dev->coherent_dma_mask;

Does setting the dma_mask pointer to the parent's coherent_dma_mask introduce
a data race?

The MFD core in mfd_add_device() copies the parent's dma_mask pointer to all
child platform devices. Because the parent's dma_mask now points to its own
dev->coherent_dma_mask field, all child devices will share the exact same
memory pointer.

During child probe, the platform bus calls of_dma_configure_id(), which
executes an unsynchronized read-modify-write on that shared pointer:

    *dev->dma_mask &= mask;

Because sec_pmic_probe() supports devices like the S2MPS11, whose regulator
driver uses PROBE_PREFER_ASYNCHRONOUS, child devices can probe concurrently.
Will these simultaneous read-modify-writes to the parent's coherent_dma_mask
cause KCSAN warnings about data races?

Additionally, does manually mutating bus-managed device properties to simulate
a 0-bit DMA capability circumvent standard DMA API validation?

> +
>  	pm_runtime_set_active(sec_pmic->dev);

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260515-s2mu005-pmic-v6-0-1979106992d4@disroot.org?part=5