Message-ID: <6683bd67-1f81-47f7-88c1-bdbf9a589bbf@oss.qualcomm.com>
Date: Thu, 14 Aug 2025 11:22:31 +0200
From: Konrad Dybcio <konrad.dybcio@....qualcomm.com>
To: Stephan Gerhold <stephan.gerhold@...aro.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
Cc: Bjorn Andersson <andersson@...nel.org>,
        Konrad Dybcio <konradybcio@...nel.org>, linux-arm-msm@...r.kernel.org,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org,
        Abel Vesa <abel.vesa@...aro.org>, Johan Hovold <johan@...nel.org>
Subject: Re: [PATCH 0/9] arm64: dts: qcom: x1: Disable audio codecs by default

On 8/14/25 10:50 AM, Stephan Gerhold wrote:
> On Thu, Aug 14, 2025 at 08:07:17AM +0200, Krzysztof Kozlowski wrote:
>> On 14/08/2025 01:09, Konrad Dybcio wrote:
>>> On 8/13/25 5:58 PM, Stephan Gerhold wrote:
>>>> Currently, the macro audio codecs are enabled by default in x1e80100.dtsi.
>>>> However, they do not probe without the ADSP remoteproc, which is disabled
>>>> by default.
>>>
>>> FWIW if the ADSP doesn't start, you can't really consider the platform
>>> working.. It just does oversees too much of the SoC to even seriously
>>> consider using the device without it
>>
>>
>> I agree. ADSP is supposed to come up for every or almost every platform,
>> because it is crucial for USB and charging.
>>
> 
> I agree with that as well, especially because I have an upcoming patch
> series that allows reusing the "lite" ADSP firmware from UEFI for USB
> and charging, so you don't even need to have firmware present for that.

Really nice!

> The question for this patch series is separate though: Should we enable
> the SoC audio codecs by default? What happens if a board does not make
> use of them?

Then they (should) get parked (powered down, or re-programmed if
necessary)

>> It's true that LPASS macro codec nodes need resources from ADSP, but
>> still these are resources internal to the SoC. We disable nodes in DTI
>> which need an external resource. That's not really the case for LPASS.
> 
> The reason that triggered this patch series is that I was seeing an
> error from the va_macro when testing on x1e001de-devkit. That board does
> not have DMICs defined, so it doesn't make direct use of the va_macro:
> 
>  va_macro 6d44000.codec: qcom,dmic-sample-rate dt entry missing

Perhaps we can print the error if there's any sound connections, i.e.
if it's "really" used?

> We should fix this in the lpass-va-macro driver. You could take this
> case one step further though: What if a board uses none of the audio
> functionality? Apparently, X1E is also going to be an IoT platform. It's
> very well possible we will end up with a board that doesn't have any
> audio functionality. I would argue it's valid to use a minimal kernel
> config in that case that has all of the audio subsystem disabled. That
> won't work though, since we need to probe all the enabled audio codecs
> to reach sync_state().

Because of the resources being driven by the OS, I don't think removing
information from the device tree (i.e. cutting down on the plumbing
data) is fair.. Enabling all the hardware (minus firmwares) should result
in only a couple hundred kilobytes of added RAM use, but will get rid of
a huge number of edge cases where sketchy combinations cause annoying
issues.

> This might be a Linux issue unrelated to the device tree, but in my
> opinion an audio codec without audio inputs/outputs is not
> "operational", it should not be status = "okay". That's quite subjective
> though.

If the codec is present on board, there's no less reason to disable it
vs leaving it hanging, unaware by the OS.. cutting the power by hand is
at least predictable

Konrad

> At the end, I realized that x1e001de-devkit actually does have DMICs and
> I just need to enable them properly to get rid of the error. I only sent
> this series because I believe it fits better to our conventions. Given
> that I don't need it anymore, I'm also happy to just drop it. Let me
> know what you prefer.
> 
> Thanks,
> Stephan

Powered by blists - more mailing lists