RE: [PATCH 2/2] ASoC: sdw_utils: skip unattached SoundWire peripherals

[Liao, Bard]

> -----Original Message-----
> From: Amaan Lalani <zlzzm2014@xxxxxxxxxxx>
> Sent: Saturday, August 29, 2026 3:04 AM
> To: Mark Brown <broonie@xxxxxxxxxx>
> Cc: Liam Girdwood <lgirdwood@xxxxxxxxx>; Jaroslav Kysela
> <perex@xxxxxxxx>; Takashi Iwai <tiwai@xxxxxxxx>; Peter Ujfalusi
> <peter.ujfalusi@xxxxxxxxxxxxxxx>; Bard Liao <yung-
> chuan.liao@xxxxxxxxxxxxxxx>; Pierre-Louis Bossart <pierre-
> louis.bossart@xxxxxxxxx>; linux-sound@xxxxxxxxxxxxxxx; sound-open-
> firmware@xxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Amaan Lalani
> <zlzzm2014@xxxxxxxxxxx>
> Subject: [PATCH 2/2] ASoC: sdw_utils: skip unattached SoundWire peripherals
> 
> Firmware may describe SoundWire peripherals which do not physically
> enumerate on the bus. The corresponding SoundWire device can still be
> present with SDW_SLAVE_UNATTACHED status.
> 
> Counting and parsing endpoints for such peripherals can create entries
> for devices which are not actually present and therefore interfere with
> sound card registration.
> 
> Check the SoundWire slave status before counting or parsing a
> peripheral. Skip peripherals reported as SDW_SLAVE_UNATTACHED while
> preserving the existing behavior when the SoundWire device cannot be
> looked up.
> 
> Fixes: 27fd36aefa00 ("ASoC: Intel: sof-sdw: Add new code for parsing the
> snd_soc_acpi structs")
> 
> Assisted-by: LLM
> Signed-off-by: Amaan Lalani <zlzzm2014@xxxxxxxxxxx>
> ---
>  sound/soc/sdw_utils/soc_sdw_utils.c | 35
> +++++++++++++++++++++++++++++
>  1 file changed, 35 insertions(+)
> 
> diff --git a/sound/soc/sdw_utils/soc_sdw_utils.c
> b/sound/soc/sdw_utils/soc_sdw_utils.c
> index 8a07ba2a29e5..381eb884efb9 100644
> --- a/sound/soc/sdw_utils/soc_sdw_utils.c
> +++ b/sound/soc/sdw_utils/soc_sdw_utils.c
> @@ -1854,6 +1854,33 @@ static int is_sdca_aux_dev_present(struct device
> *dev,
>  	return 0;
>  }
> 
> +static bool is_peripheral_attached(struct device *dev,
> +				    const struct snd_soc_acpi_link_adr
> *adr_link,
> +				    int adr_index)
> +{
> +	const char *sdw_codec_name;
> +	struct sdw_slave *slave;
> +
> +	sdw_codec_name = _asoc_sdw_get_codec_name(dev, adr_link,
> adr_index);
> +	if (!sdw_codec_name)
> +		return true;
> +
> +	struct device *sdw_dev __free(put_device) =
> +	  bus_find_device_by_name(&sdw_bus_type, NULL,
> sdw_codec_name);
> +	if (!sdw_dev)
> +		return true;
> +
> +	slave = dev_to_sdw_dev(sdw_dev);
> +
> +	if (slave->status == SDW_SLAVE_UNATTACHED) {
> +		dev_dbg(dev, "%s not present on the bus, skipping\n",
> +		  sdw_codec_name);
> +		return false;
> +	}
> +
> +	return true;
> +}
> +
>  int asoc_sdw_count_sdw_endpoints(struct snd_soc_card *card,
>  				 int *num_devs, int *num_ends, int
> *num_aux)
>  {
> @@ -1870,6 +1897,11 @@ int asoc_sdw_count_sdw_endpoints(struct
> snd_soc_card *card,
>  			const struct snd_soc_acpi_adr_device *adr_dev =
> &adr_link->adr_d[i];
>  			struct asoc_sdw_codec_info *codec_info;
> 
> +			if (!is_peripheral_attached(dev, adr_link, i)) {

Sorry, but I don't think it is correct.
The status just means the current state. A Peripheral aka Slave on the
bus could be attached or unattached. We can't use the slave->status to
determine whether a Peripheral is physically on the bus or not.
Checking slave->dev_num_sticky may work. However, there is a timing
issue that the Peripheral could be attached after the check.

> +				(*num_devs)--;
> +				continue;
> +			}
> +
>  			*num_ends += adr_dev->num_endpoints;
> 
>  			codec_info =
> asoc_sdw_find_codec_info_part(adr_dev->adr);
> @@ -2029,6 +2061,9 @@ int asoc_sdw_parse_sdw_endpoints(struct device
> *dev,
>  				return -EINVAL;
>  			}
> 
> +			if (!is_peripheral_attached(dev, adr_link, i))
> +				continue;
> +
>  			codec_info =
> asoc_sdw_find_codec_info_part(adr_dev->adr);
>  			if (!codec_info)
>  				return -EINVAL;
> --
> 2.55.0
>