[bug report] 'ASoC: Intel: haswell: Power transition refactor' and PulseAudio

[bug report] 'ASoC: Intel: haswell: Power transition refactor' and PulseAudio

[Christian Bundy]

After upgrading to Linux 5.8 I discovered an audio issue on my device that was introduced in 8ec7d6043263ecf250b9b7c0dd8ade899487538a [0]. I used 'git bisect' to identify the commit that introduced the bug and have confirmed that reverting the commit resolves the problem

Reproduction:

1. Play any audio via PulseAudio.
2. Observe that the audio output is fuzzy and choppy.

I can use programs like mpv to play audio without PulseAudio, and the audio is fine, but as soon as I open a process that uses PulseAudio it will ruin the audio output for all processes (including mpv) until I reboot.

I'm using a 2015 Chromebook Pixel ("Samus") and have confirmed this problem with a friend who has the same device.

Is there anything I can do to help debug this instead of sending a patch to revert the commit?

Relevant lspci output:

    00:03.0 Audio device: Intel Corporation Broadwell-U Audio Controller (rev 09)
	Subsystem: Intel Corporation Broadwell-U Audio Controller
	Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx+
	Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
	Latency: 0, Cache Line Size: 64 bytes
	Interrupt: pin A routed to IRQ 50
	Region 0: Memory at e1218000 (64-bit, non-prefetchable) [size=16K]
	Capabilities: [50] Power Management version 2
		Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
		Status: D0 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME-
	Capabilities: [60] MSI: Enable+ Count=1/1 Maskable- 64bit-
		Address: fee00378  Data: 0000
	Capabilities: [70] Express (v1) Root Complex Integrated Endpoint, MSI 00
		DevCap:	MaxPayload 128 bytes, PhantFunc 0
			ExtTag- RBE- FLReset+
		DevCtl:	CorrErr- NonFatalErr- FatalErr- UnsupReq-
			RlxdOrd- ExtTag- PhantFunc- AuxPwr- NoSnoop- FLReset-
			MaxPayload 128 bytes, MaxReadReq 128 bytes
		DevSta:	CorrErr- NonFatalErr- FatalErr- UnsupReq- AuxPwr- TransPend-
	Kernel driver in use: snd_hda_intel
	Kernel modules: snd_hda_intel

[0]: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=8ec7d6043263ecf250b9b7c0dd8ade899487538a

Re: [bug report] 'ASoC: Intel: haswell: Power transition refactor' and PulseAudio

[Cezary Rojewski]

On 2020-08-31 11:55 PM, Christian Bundy wrote:
> After upgrading to Linux 5.8 I discovered an audio issue on my device that was introduced in 8ec7d6043263ecf250b9b7c0dd8ade899487538a [0]. I used 'git bisect' to identify the commit that introduced the bug and have confirmed that reverting the commit resolves the problem
> 
> Reproduction:
> 
> 1. Play any audio via PulseAudio.
> 2. Observe that the audio output is fuzzy and choppy.
> 
> I can use programs like mpv to play audio without PulseAudio, and the audio is fine, but as soon as I open a process that uses PulseAudio it will ruin the audio output for all processes (including mpv) until I reboot.
> 
> I'm using a 2015 Chromebook Pixel ("Samus") and have confirmed this problem with a friend who has the same device.
> 
> Is there anything I can do to help debug this instead of sending a patch to revert the commit?
> 

Hello Christian,

Thank you for report! Issue is a known one to us and has already been 
addressed by:

	[PATCH v4 00/13] ASoC: Intel: Catpt - Lynx and Wildcat point
	https://www.spinics.net/lists/alsa-devel/msg113762.html

waiting for final dependency to be merged (Andy's resource-API changes, 
as Mark already added the SPI ones) so v5 with review changes can be 
provided. Shouldn't be long before this gets merged. As consequence, 
/haswell/ ceases to exist.

Basically, once power-cycle (D0 -> D3 -> D0 transition flow) had been 
fixed, more - previously hidden - problems arisen. Instead of sending 
70+ patches to Mark refactoring existing code to recommended flow (+ 
readability and performance improvements), replacement is provided along 
with old code being removed entirely.

For now, if there's a possibility for you to modify your kernel, said 
patch can be safely removed from your local repo. Note: following is the 
outcome:
- DMA init may occasionally fail on early boot (audio card won't be 
present at all, requires reboot)
- D0/D3 flow doesn't follow recommended sequence and thus power-saving 
may be limited or non-existent
Probably still better than permanently fuzzied audio..

I'm sorry for any inconvenience this has caused to you.

Regards,
Czarek

Re: [bug report] 'ASoC: Intel: haswell: Power transition refactor' and PulseAudio

[Pierre-Louis Bossart]

On 9/1/20 6:33 AM, Cezary Rojewski wrote:
> On 2020-08-31 11:55 PM, Christian Bundy wrote:
>> After upgrading to Linux 5.8 I discovered an audio issue on my device 
>> that was introduced in 8ec7d6043263ecf250b9b7c0dd8ade899487538a [0]. I 
>> used 'git bisect' to identify the commit that introduced the bug and 
>> have confirmed that reverting the commit resolves the problem
>>
>> Reproduction:
>>
>> 1. Play any audio via PulseAudio.
>> 2. Observe that the audio output is fuzzy and choppy.
>>
>> I can use programs like mpv to play audio without PulseAudio, and the 
>> audio is fine, but as soon as I open a process that uses PulseAudio it 
>> will ruin the audio output for all processes (including mpv) until I 
>> reboot.
>>
>> I'm using a 2015 Chromebook Pixel ("Samus") and have confirmed this 
>> problem with a friend who has the same device.
>>
>> Is there anything I can do to help debug this instead of sending a 
>> patch to revert the commit?
>>
> 
> Hello Christian,
> 
> Thank you for report! Issue is a known one to us and has already been 
> addressed by:
> 
>      [PATCH v4 00/13] ASoC: Intel: Catpt - Lynx and Wildcat point
>      https://www.spinics.net/lists/alsa-devel/msg113762.html
> 
> waiting for final dependency to be merged (Andy's resource-API changes, 
> as Mark already added the SPI ones) so v5 with review changes can be 
> provided. Shouldn't be long before this gets merged. As consequence, 
> /haswell/ ceases to exist.

That leaves people with no working sound for 5.8 and 5.9.

> Basically, once power-cycle (D0 -> D3 -> D0 transition flow) had been 
> fixed, more - previously hidden - problems arisen. Instead of sending 
> 70+ patches to Mark refactoring existing code to recommended flow (+ 
> readability and performance improvements), replacement is provided along 
> with old code being removed entirely.
> 
> For now, if there's a possibility for you to modify your kernel, said 
> patch can be safely removed from your local repo. Note: following is the 
> outcome:
> - DMA init may occasionally fail on early boot (audio card won't be 
> present at all, requires reboot)
> - D0/D3 flow doesn't follow recommended sequence and thus power-saving 
> may be limited or non-existent
> Probably still better than permanently fuzzied audio..

Doesn't this mean that a revert is needed and applied to -stable for 5.8 
and 5.9?

Re: [bug report] 'ASoC: Intel: haswell: Power transition refactor' and PulseAudio

[Cezary Rojewski]

On 2020-09-01 3:38 PM, Pierre-Louis Bossart wrote:
> On 9/1/20 6:33 AM, Cezary Rojewski wrote:

...

>>
>> Hello Christian,
>>
>> Thank you for report! Issue is a known one to us and has already been 
>> addressed by:
>>
>>      [PATCH v4 00/13] ASoC: Intel: Catpt - Lynx and Wildcat point
>>      https://www.spinics.net/lists/alsa-devel/msg113762.html
>>
>> waiting for final dependency to be merged (Andy's resource-API 
>> changes, as Mark already added the SPI ones) so v5 with review changes 
>> can be provided. Shouldn't be long before this gets merged. As 
>> consequence, /haswell/ ceases to exist.
> 
> That leaves people with no working sound for 5.8 and 5.9.
> 
>> Basically, once power-cycle (D0 -> D3 -> D0 transition flow) had been 
>> fixed, more - previously hidden - problems arisen. Instead of sending 
>> 70+ patches to Mark refactoring existing code to recommended flow (+ 
>> readability and performance improvements), replacement is provided 
>> along with old code being removed entirely.
>>
>> For now, if there's a possibility for you to modify your kernel, said 
>> patch can be safely removed from your local repo. Note: following is 
>> the outcome:
>> - DMA init may occasionally fail on early boot (audio card won't be 
>> present at all, requires reboot)
>> - D0/D3 flow doesn't follow recommended sequence and thus power-saving 
>> may be limited or non-existent
>> Probably still better than permanently fuzzied audio..
> 
> Doesn't this mean that a revert is needed and applied to -stable for 5.8 
> and 5.9?
> 

I believe you're right Pierre, revert should be provided. I'll see to it.

Czarek

Re: [bug report] 'ASoC: Intel: haswell: Power transition refactor' and PulseAudio

[Liam Girdwood]

On Tue, 2020-09-01 at 13:33 +0200, Cezary Rojewski wrote:
> On 2020-08-31 11:55 PM, Christian Bundy wrote:
> > After upgrading to Linux 5.8 I discovered an audio issue on my
> > device that was introduced in
> > 8ec7d6043263ecf250b9b7c0dd8ade899487538a [0]. I used 'git bisect'
> > to identify the commit that introduced the bug and have confirmed
> > that reverting the commit resolves the problem
> > 
> > Reproduction:
> > 
> > 1. Play any audio via PulseAudio.
> > 2. Observe that the audio output is fuzzy and choppy.
> > 
> > I can use programs like mpv to play audio without PulseAudio, and
> > the audio is fine, but as soon as I open a process that uses
> > PulseAudio it will ruin the audio output for all processes
> > (including mpv) until I reboot.
> > 
> > I'm using a 2015 Chromebook Pixel ("Samus") and have confirmed this
> > problem with a friend who has the same device.
> > 
> > Is there anything I can do to help debug this instead of sending a
> > patch to revert the commit?
> > 
> 
> Hello Christian,
> 
> Thank you for report! Issue is a known one to us and has already
> been 
> addressed by:
> 
> 	[PATCH v4 00/13] ASoC: Intel: Catpt - Lynx and Wildcat point
> 	https://www.spinics.net/lists/alsa-devel/msg113762.html
> 
> waiting for final dependency to be merged (Andy's resource-API
> changes, 
> as Mark already added the SPI ones) so v5 with review changes can be 
> provided. Shouldn't be long before this gets merged. As consequence, 
> /haswell/ ceases to exist.
> 

Please also don't forget that the new BDW HW register programming flows
need to be shared as common code with the SOF BDW driver.

Thanks

Liam

Re: [bug report] 'ASoC: Intel: haswell: Power transition refactor' and PulseAudio

[Cezary Rojewski]

On 2020-09-01 5:23 PM, Liam Girdwood wrote:
> On Tue, 2020-09-01 at 13:33 +0200, Cezary Rojewski wrote:

...

>>
>> Hello Christian,
>>
>> Thank you for report! Issue is a known one to us and has already
>> been
>> addressed by:
>>
>> 	[PATCH v4 00/13] ASoC: Intel: Catpt - Lynx and Wildcat point
>> 	https://www.spinics.net/lists/alsa-devel/msg113762.html
>>
>> waiting for final dependency to be merged (Andy's resource-API
>> changes,
>> as Mark already added the SPI ones) so v5 with review changes can be
>> provided. Shouldn't be long before this gets merged. As consequence,
>> /haswell/ ceases to exist.
>>
> 
> Please also don't forget that the new BDW HW register programming flows
> need to be shared as common code with the SOF BDW driver.
> 
> Thanks
> 
> Liam
> 

I don't believe this is related to Christian's report.

Anyway, revert-patch for sound/soc/intel/haswell/ solution has been 
provided so it can be later propagated to v5.8 and v5.9 -stable with 
matching upstream commit's sha-id.

Czarek

Re: [bug report] 'ASoC: Intel: haswell: Power transition refactor' and PulseAudio

[Christian Bundy]

Thanks for the quick response. This is great. I've tested the patch and confirmed that it works.

On Tue, Sep 1, 2020, at 08:37, Cezary Rojewski wrote:
> On 2020-09-01 5:23 PM, Liam Girdwood wrote:
> > On Tue, 2020-09-01 at 13:33 +0200, Cezary Rojewski wrote:
> 
> ...
> 
> >>
> >> Hello Christian,
> >>
> >> Thank you for report! Issue is a known one to us and has already
> >> been
> >> addressed by:
> >>
> >> 	[PATCH v4 00/13] ASoC: Intel: Catpt - Lynx and Wildcat point
> >> 	https://www.spinics.net/lists/alsa-devel/msg113762.html
> >>
> >> waiting for final dependency to be merged (Andy's resource-API
> >> changes,
> >> as Mark already added the SPI ones) so v5 with review changes can be
> >> provided. Shouldn't be long before this gets merged. As consequence,
> >> /haswell/ ceases to exist.
> >>
> > 
> > Please also don't forget that the new BDW HW register programming flows
> > need to be shared as common code with the SOF BDW driver.
> > 
> > Thanks
> > 
> > Liam
> > 
> 
> I don't believe this is related to Christian's report.
> 
> Anyway, revert-patch for sound/soc/intel/haswell/ solution has been 
> provided so it can be later propagated to v5.8 and v5.9 -stable with 
> matching upstream commit's sha-id.
> 
> Czarek
>

Re: [bug report] 'ASoC: Intel: haswell: Power transition refactor' and PulseAudio

[Liam Girdwood]

On Tue, 2020-09-01 at 17:37 +0200, Cezary Rojewski wrote:
> > > Thank you for report! Issue is a known one to us and has already
> > > been
> > > addressed by:
> > >       [PATCH v4 00/13] ASoC: Intel: Catpt - Lynx and Wildcat
> > > point
> > >       https://www.spinics.net/lists/alsa-devel/msg113762.html
> > > waiting for final dependency to be merged (Andy's resource-API
> > > changes,
> > > as Mark already added the SPI ones) so v5 with review changes can
> > > be
> > > provided. Shouldn't be long before this gets merged. As
> > > consequence,
> > > /haswell/ ceases to exist.
> > Please also don't forget that the new BDW HW register programming
> > flows
> > need to be shared as common code with the SOF BDW driver.
> > Thanks
> > Liam
> 
> 
> I don't believe this is related to Christian's report.

To be clear, it's related to v5 readiness statement and not this
report.

Thanks

Liam