Oneiric: [PATCH] (UBUNTU: Sauce) Match jack input devices and pcm devices for HDMI

[David Henningsson]

On 08/31/2011 11:35 PM, Leann Ogasawara wrote:
> On Wed, 2011-08-31 at 09:18 +0200, David Henningsson wrote:
>> On 08/30/2011 03:48 PM, Tim Gardner wrote:
> [...]
>>> So, you're asking to have a patch applied that has general impact on all
>>> platforms with an HDMI port, which by your own admission has not had
>>> widespread testing, and has not been agreed to or fully reviewed by
>>> upstream maintainers.
>>>
>>> The kernel is in Beta freeze right now. What would your call be if you
>>> were in the same position?
>>
>> If I were in the same position:
>>
>> First and foremost, I would start with a warm, welcoming and encouraging
>> attitude towards patches. A patch sent here, means someone not only
>> wants to improve Linux in general, but also cares in special for Ubuntu,
>> and has taken the time and effort to backport a patch to suit Ubuntu's
>> particular kernel version. In this case, the patch author is a Canonical
>> employee so this is not surprising, but nevertheless, this is a really
>> good thing and should be encouraged.
>
> Let me just start by saying that we really appreciate all the work that
> you've done with regards to the audio subsystem.  Your interaction with
> upstream and the bug fixes you've provided for Ubuntu have been much
> appreciated and I encourage you to keep up the good work.
>
>> With that positive attitude in mind, I would do a review. The review
>> would, of course, take regression risk into account - a kernel bug can
>> cause the system to fail in the most horrifying ways - but also weigh
>> that against the possible positive outcomes of applying the patch, i e
>> why it's needed in the first place. If I'm not qualified to make a
>> review, I would find someone who is, to make the opinion for me.
>
> Because we do consider you one of our knowledge experts in the audio
> area, I believe Tim was wanting a bit more reassurance and confidence
> from you with regards to the patch.  Having recognized the potential for
> widespread impact, I do feel he was doing his job as one of the
> maintainers of the Ubuntu kernel to make sure we're not introducing
> regressions, especially being so close to Beta for Oneiric.
>
>> The review in this case would notice that it impacts most people with
>> working HDMI audio which means both higher risk and gain. OTOH, the code
>> path is simple, which means that successful testing on one machine would
>> make it unlikely to fail on another.
>>
>> If I'm still unsure of regressions even after having done the review, I
>> would make sure testing of the patch is done. If I have time, I do the
>> testing myself, or if I don't, ask for someone to test it for me, but do
>> what I can to help, e g by building a test kernel and provide
>> instructions as of how to do the test. If testing with positive result
>> is all that's needed before applying the patch, I would communicate that
>> clearly, among with a deadline explaining when the testing must be
>> completed.
>
> I've built a test kernel with your patch applied and placed it at the
> following location:
>
> http://people.canonical.com/~ogasawara/diwic/
>
> If you can, could you please test and confirm this resolves the issue at
> hand and also doesn't introduce any regressions on a few existing
> systems with working HDMI audio (whatever access to systems you have on
> hand would suffice).  Please keep in mind that Oneiric Kernel Freeze is
> Thurs Sept 15, so if I could get your feedback before then I'd be happy
> to try and get this applied before we hit the freeze date.

Hi Leann and thank you for your kind words and clear instructions! :-)

Here are the testing results (where 3.0.0-10 means your patched and 
built test kernel installed, and 3.0.0-9 is what comes with 11.10 beta 1) :

Radeon HD 4200 series:

3.0.0-9, radeon driver: no jack input device, no working HDMI audio.
3.0.0-10, radeon driver: no jack input device, no working HDMI audio.
3.0.0-9, fglrx driver: no jack input device, working HDMI audio.
3.0.0-10, fglrx driver: no jack input device, working HDMI audio.

Note: Since there is no jack input device, the code path is not run and 
my patch does not make any difference. Might look into that later (why 
there isn't a jack input device), but that is not on top of the priority 
list ATM.

Intel Arrandale:

3.0.0-9: one jack input device "HDA Intel HDMI/DP" (plugin working, 
unplug does not work), working HDMI audio.
3.0.0-10: one jack input device "HDA Intel HDMI/DP,pcm=3" (plugin 
working but compiz crashed at one point, unplug does not work), working 
HDMI audio.

Note: Patch working as jack input device name changes successfully. The 
compiz crash (reported as bug 839468) was a little strange, so I retried 
but could not reproduce the crash under either kernel. I don't think it 
has anything to do with the HDMI audio patch.

Nvidia GT 430:

3.0.0-9, nouveau driver: four jack input devices "HDA NVidia HDMI/DP" 
(none of them working), no working HDMI audio on any of the four PCMs.
3.0.0-10, nouveau driver: four jack input devices "HDA NVidia 
HDMI/DP,pcm=3", "HDA NVidia HDMI/DP,pcm=7", "HDA NVidia HDMI/DP,pcm=8", 
"HDA NVidia HDMI/DP,pcm=9" (none of them working), no working HDMI audio 
on any of the four PCMs.
3.0.0-10, nvidia driver: four jack input devices "HDA NVidia 
HDMI/DP,pcm=3", "HDA NVidia HDMI/DP,pcm=7", "HDA NVidia HDMI/DP,pcm=8", 
"HDA NVidia HDMI/DP,pcm=9". The pcm=9 indicates working HDMI audio for 
plughw:NVidia,9 (or hdmi:NVidia,3). \o/

Note: I spent four hours trying to install nvidia binary driver before I 
succeeded (thanks to Sarvatt, tjaalton, and tseliot for helping out), 
and one more hour trying to install it under 3.0.0.9.

Final conclusion: patch works as expected and no regressions found.

-- 
David Henningsson, Canonical Ltd.
http://launchpad.net/~diwic