Re: [PATCH 3/4] ASoC: fsl_asrc: expose individual DAIs per conversion pair

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] The newly registered DAIs (`paira`, `pairb`, `pairc`) share the exact same `dma_params_tx` and `dma_params_rx` structures, causing a data race and potential DMA misconfiguration when multiple streams are configured concurrently.
--

--- Patch [3]: [PATCH 3/4] ASoC: fsl_asrc: expose individual DAIs per conversion pair ---
commit 71178f6edf2667b8ee88b95ef47b739e170cca4b
Author: Shengjiu Wang <shengjiu.wang@nxp.com>

ASoC: fsl_asrc: expose individual DAIs per conversion pair

The i.MX ASRC hardware supports three independent conversion pairs
(A, B, C). This commit replaces the single fsl_asrc_dai instance
with an array of three DAI drivers, one per pair, allowing them
to be routed to separate audio paths simultaneously.

> diff --git a/sound/soc/fsl/fsl_asrc.c b/sound/soc/fsl/fsl_asrc.c
> index 0b28bcfa47fed..f83d6cdc6412d 100644
> --- a/sound/soc/fsl/fsl_asrc.c
> +++ b/sound/soc/fsl/fsl_asrc.c

[ ... ]

> @@ -1404,7 +1453,7 @@ static int fsl_asrc_probe(struct platform_device *pdev)
>  		goto err_pm_get_sync;
>  
>  	ret = devm_snd_soc_register_component(&pdev->dev, &fsl_asrc_component,
> -					      &fsl_asrc_dai, 1);
> +					      fsl_asrc_dai, ARRAY_SIZE(fsl_asrc_dai));

[Severity: High]
Since this registers multiple DAIs to enable concurrent stream access, can this
cause a data race with the shared DMA parameters?

In fsl_asrc_dai_probe(), all three DAIs are assigned pointers to the exact same
shared DMA parameter structures:

	snd_soc_dai_init_dma_data(dai, &asrc->dma_params_tx,
				  &asrc->dma_params_rx);

When multiple streams are configured concurrently, they will race to overwrite
these shared parameters without locking.

For example, in fsl_asrc_dma_hw_params():

	dma_params_fe = snd_soc_dai_get_dma_data(snd_soc_rtd_to_cpu(rtd, 0), substream);
	dma_params_fe->addr = asrc->paddr + asrc->get_fifo_addr(!dir, index);
	dma_params_fe->maxburst = dma_params_be->maxburst;

Could this concurrent mutation cause DMA channels to be programmed with incorrect
FIFO addresses or maxburst values, resulting in audio cross-talk or DMA engine
errors?

>  	if (ret) {
>  		dev_err(&pdev->dev, "failed to register ASoC DAI\n");
>  		goto err_pm_get_sync;

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260720095450.1618747-1-shengjiu.wang@oss.nxp.com?part=3