Re: [PATCH v2] ASoC: codecs: wcd938x: fix accessing regmap on unattached devices

[Pierre-Louis Bossart]

On 5/3/23 09:41, Krzysztof Kozlowski wrote:
> The WCD938x comes with three devices on two Linux drivers:
> 1. RX Soundwire device (wcd938x-sdw.c driver),
> 2. TX Soundwire device, which is used to access devices via regmap (also
>    wcd938x-sdw.c driver),
> 3. platform device (wcd938x.c driver) - glue and component master,
>    actually having most of the code using TX Soundwire device regmap.
> 
> When RX and TX Soundwire devices probe, the component master (platform
> device) bind tries to write micbias configuration via TX Soundwire
> regmap.  This might happen before TX Soundwire enumerates, so the regmap
> access fails.  On Qualcomm SM8550 board with WCD9385:
> 
>   qcom-soundwire 6d30000.soundwire-controller: Qualcomm Soundwire controller v2.0.0 Registered
>   wcd938x_codec audio-codec: bound sdw:0:0217:010d:00:4 (ops wcd938x_sdw_component_ops)
>   wcd938x_codec audio-codec: bound sdw:0:0217:010d:00:3 (ops wcd938x_sdw_component_ops)
>   qcom-soundwire 6ad0000.soundwire-controller: swrm_wait_for_wr_fifo_avail err write overflow
> 
> Fix the issue by:
> 1. Moving the regmap creation from platform device to TX Soundwire
>    device.  The regmap settings are moved as-is with one difference:
>    making the wcd938x_regmap_config const.
> 2. Using regmap in cache only mode till the actual TX Soundwire device
>    enumerates and then sync the regmap cache.
> 
> Cc: <stable@xxxxxxxxxxxxxxx> # v3.14+
> Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx>
> 
> ---
> 
> Cc: Pierre-Louis Bossart <pierre-louis.bossart@xxxxxxxxxxxxxxx>
> Cc: Patrick Lai <quic_plai@xxxxxxxxxxx>
> Cc: Srinivas Kandagatla <srinivas.kandagatla@xxxxxxxxxx>
> 
> Changes since v1:
> 1. Previous solution touched Soundwire. Rework it according to feedback.
>    https://lore.kernel.org/all/20230420101617.142225-7-krzysztof.kozlowski@xxxxxxxxxx/
> 
> Thank you Mark and Pierre-Louis for great hints and valuable feedback!

LGTM, this was a very interesting thread to comment on.

Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@xxxxxxxxxxxxxxx>