Re: [PATCH v2 10/11] ASoC: qcom: Add Shikra QAIF support

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] Missing DAI descriptors for CDC DMA TX (capture), RX2-3, and VA_TX2-3 channels render hardware paths completely inaccessible.
- [Low] Monolithic module linking prevents addition of future SoC variants without build errors.
--

commit 8fb121461863147da2f4ca2b57d93d08f040ef59
Author: Harendra Gautam <harendra.gautam@oss.qualcomm.com>
Subject: ASoC: qcom: Add Shikra QAIF support

This commit adds the Shikra variant data for the Qualcomm Audio Interface
driver. It introduces register field definitions, DMA mappings, DAI descriptors,
and clock configuration. The new variant is hooked up to Kconfig and the qcom
ASoC Makefile.

> diff --git a/sound/soc/qcom/Makefile b/sound/soc/qcom/Makefile
> index 985ce2ae286ba..857bb2a032a2d 100644
> --- a/sound/soc/qcom/Makefile
> +++ b/sound/soc/qcom/Makefile
> @@ -31,6 +31,7 @@ snd-soc-qcom-common-y := common.o
>  snd-soc-qcom-sdw-y := sdw.o
>  snd-soc-x1e80100-y := x1e80100.o
>  snd-soc-qcom-offload-utils-objs := usb_offload_utils.o
> +snd-soc-qcom-qaif-y := qaif-cpu.o qaif-platform.o qaif-shikra.o

[ ... ]

> diff --git a/sound/soc/qcom/qaif-shikra.c b/sound/soc/qcom/qaif-shikra.c
> new file mode 100644
> index 0000000000000..6bf32d155f3a1
> --- /dev/null
> +++ b/sound/soc/qcom/qaif-shikra.c
> @@ -0,0 +1,556 @@

[ ... ]

> +static const struct qaif_dmaidx_dai_map shikra_cif_tx_dma_dai_map[] = {
> +		{ QAIF_CDC_DMA_TX0 },
> +		{ QAIF_CDC_DMA_TX1 },
> +		{ QAIF_CDC_DMA_TX2 },
> +		{ QAIF_CDC_DMA_TX3 }
> +};
> +
> +static const struct qaif_dmaidx_dai_map shikra_cif_va_dma_dai_map[] = {
> +		{ QAIF_CDC_DMA_VA_TX0 },
> +		{ QAIF_CDC_DMA_VA_TX1 },
> +		{ QAIF_CDC_DMA_VA_TX2 },
> +		{ QAIF_CDC_DMA_VA_TX3 }
> +};

[ ... ]

> +static struct snd_soc_dai_driver shikra_qaif_cpu_dai_driver[] = {
> +	{
> +		.id = QAIF_MI2S_TDM_AIF0,

[ ... ]

> +	}, {
> +		.id = QAIF_CDC_DMA_VA_TX1,
> +		.name = "CDC DMA VA1",
> +		.capture = {
> +			.stream_name = "DMIC Capture1",
> +			.formats = SNDRV_PCM_FMTBIT_S16,
> +			.rates = SNDRV_PCM_RATE_48000,
> +			.rate_min	= 48000,
> +			.rate_max	= 48000,
> +			.channels_min	= 1,
> +			.channels_max	= 4,
> +		},
> +		.ops	= &asoc_qcom_qaif_cif_dai_ops,
> +	},
> +};

[Severity: High]
The DMA mapping (shikra_cif_tx_dma_dai_map) specifies all 4 QAIF_CDC_DMA_TX
channels, but the shikra_qaif_cpu_dai_driver array seems to omit the WCD
capture (QAIF_CDC_DMA_TX) DAIs entirely. It also only provides 2 out of 4 for
the RX and VA_TX DAIs.

Will this cause sound card registration to fail when a machine driver
attempts to instantiate a PCM capture stream routing through
QAIF_CDC_DMA_TX0 (or TX1-3)?

[ ... ]

> +static struct platform_driver shikra_qaif_cpu_platform_driver = {
> +	.driver = {
> +		.name = "shikra-qaif-cpu",
> +		.of_match_table = shikra_qaif_cpu_device_id,
> +		.pm = &shikra_qaif_pm_ops,
> +	},
> +	.probe = asoc_qcom_qaif_cpu_platform_probe,
> +	.remove = asoc_qcom_qaif_cpu_platform_remove,
> +	.shutdown = asoc_qcom_qaif_cpu_platform_shutdown,
> +};
> +module_platform_driver(shikra_qaif_cpu_platform_driver);

[Severity: Low]
Because qaif-cpu.o, qaif-platform.o, and qaif-shikra.o are linked into a
single snd-soc-qcom-qaif.ko module, does using module_platform_driver()
here inadvertently claim the entire module's init/exit entry points?

Could this lead to build failures due to multiple init_module definitions
if a future SoC variant is added to snd-soc-qcom-qaif-y and also attempts
to register a platform driver?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260701105757.2779738-1-harendra.gautam@oss.qualcomm.com?part=10