Re: [PATCH v1 3/4] ASoC: codecs: pm4125: Keep the micbias reference count balanced

[Mohammad Rafi Shaik <mohammad.rafi.shaik@oss.qualcomm.com>]

On 10/16/2026 12:33 AM, Mark Brown wrote:
> On Thu, Oct 15, 2026 at 07:43:11PM +0530, Mohammad Rafi Shaik wrote:
> 
>> The analog master bias is shared between the microphone bias supplies
>> and the playback and capture paths, so pm4125_global_mbias_enable() and
>> pm4125_global_mbias_disable() reference count it and only touch
>> PM4125_ANA_MBIAS_EN for the first user and the last one.
> 
>> MICB_DISABLE clamps the per-supply reference count at zero but then
>> evaluates the enable and pull-up counts regardless, so a release that
>> has no matching request still reaches pm4125_global_mbias_disable() and
>> drops a reference it never took. The MBHC framework issues exactly such
>> a release: it takes MIC_BIAS_2 once when it starts detecting a plug
>> type and releases it both at the end of the correction work and again
>> from wcd_micbias_disable(). Inserting a jack while a stream is running
>> therefore switches the analog master bias off underneath it, leaving
>> the path up but silent, and it stays silent until another stream takes
>> the first reference again.
>>
>> Return early when there is nothing to release, so the hardware is only
>> touched for a request that was actually made.
> 
> This feels like a bodge for poor reference counting elsewhere - in
> particular the MBHC seems to be fairly clearly double freeing.  The fix
> here just silently eats double frees AFAICT?

Hi Mark,

Good catch. I’ll check this and fix it in the next revision.

Sorry for the delayed response. I was away for a few days.

Best Regards.
Rafi.