Re: [RFT] Major snd_hda rewrite

On 14.01.2012 04:10, Michael Schnell wrote:
>
> On Thu, 12 Jan 2012, Yuri Pankov wrote:
>
>> On Thu, Jan 12, 2012 at 02:57:52PM +0200, Alexander Motin wrote:
>>> On 01/12/12 14:18, Yuri Pankov wrote:
>>>> On Wed, Jan 11, 2012 at 09:33:17PM +0200, Alexander Motin wrote:
>>>>> I would like request for testing of my work on further HDA sound
>>>>> driver
>>>>> improvement.
>>>> [...]
>>>>> Patch can be found here:
>>>>> http://people.freebsd.org/~mav/hda.rewrite.patch
>>>>>
>>>>> Patch was generated for 10-CURRENT, but should apply to fresh 9-STABLE
>>>>> and 8-STABLE branches also.
>>>>
>>>> Patch applied cleanly to r230008 using `svn patch`.
>>>>
>>>> hdacc0:<NVidia GT220 HDA CODEC> at cad 0 on hdac0
>>>> hdaa0:<NVidia GT220 HDA CODEC Audio Function Group> at nid 1 on hdacc0
>>>> pcm0:<NVidia GT220 HDA CODEC PCM (DisplayPort 8ch)> at nid 5 on hdaa0
>>>> hdacc1:<NVidia GT220 HDA CODEC> at cad 1 on hdac0
>>>> hdaa1:<NVidia GT220 HDA CODEC Audio Function Group> at nid 1 on hdacc1
>>>> pcm1:<NVidia GT220 HDA CODEC PCM (DisplayPort 8ch)> at nid 5 on hdaa1
>>>> hdacc2:<NVidia GT220 HDA CODEC> at cad 2 on hdac0
>>>> hdaa2:<NVidia GT220 HDA CODEC Audio Function Group> at nid 1 on hdacc2
>>>> pcm2:<NVidia GT220 HDA CODEC PCM (DisplayPort 8ch)> at nid 5 on hdaa2
>>>> hdacc3:<NVidia GT220 HDA CODEC> at cad 3 on hdac0
>>>> hdaa3:<NVidia GT220 HDA CODEC Audio Function Group> at nid 1 on hdacc3
>>>> pcm3:<NVidia GT220 HDA CODEC PCM (DisplayPort 8ch)> at nid 5 on hdaa3
>>>> hdacc4:<IDT 92HD75BX HDA CODEC> at cad 0 on hdac1
>>>> hdaa4:<IDT 92HD75BX HDA CODEC Audio Function Group> at nid 1 on hdacc4
>>>> pcm4:<IDT 92HD75BX HDA CODEC PCM (Analog)> at nid 13 and 11 on hdaa4
>>>> pcm5:<IDT 92HD75BX HDA CODEC PCM (Analog)> at nid 15 and 24 on hdaa4
>>>> pcm6:<IDT 92HD75BX HDA CODEC PCM (Front Digital)> at nid 30 on hdaa4
>>>>
>>>> pcm4 (builtin speakers) and pcm5 (headphones) seem to work fine,
>>>> however
>>>
>>> Thank you.
>>>
>>>> I'm not getting anything out of pcm0-pcm3 (connected to a TV via HDMI),
>>>> mplayer just pauses at the beggining, trying to cat anything to
>>>> /dev/dsp{0-3}.0 gives:
>>>>
>>>> pcm0: chn_write(): pcm0:virtual:dsp0.vp0: play interrupt timeout,
>>>> channel dead
>>>>
>>>> It was the same with the old driver and I'm not sure if it's (most
>>>> likely) my misconfiguration or a driver problem.
>>>
>>> It sounds more like a driver problem. HDMI audio is still not very well
>>> discovered area, and, according to ALSA reading, NVidia HDMI is also not
>>> very standard. Probably I'll finally have to buy something to
>>> experiment. What card do you have?
>>
>> It's a laptop with "nVidia Corporation GT216 [GeForce GT 230M]" (as
>> identified by x11/nvidia-driver).
>>
>> The verbose dmesg is at:
>>
>> https://www.xvoid.org/stuff/spica.dmesg
>
> I had the same problem for some time and I was going over the code and
> honestly I took some inspiration of how the linux kernel handle this and
> added some quirks to the code to get it working. However, after some
> time I realize that a sysctl
> dev.hdac.<n>.0.polling=1 will do something similar and this also works for
> me. I don't think this polling mechanism is a good idea, better would be
> some interrupt for state updates but anyway, I was glad that I can here
> (digital) music over my receiver with my laptop. I have an NVS 3100M
> graphic card and it is connected over a (fragile) Displayport-HDMI adapter.
>
> I would gladly test this with Alexanders patch, but I only have an
> 9.0-RELEASE and I already tried to port the patch back, but this would
> take some effort. In addition this is my production system and I already
> messed it up enough. ;)

I also don't like polling, especially in context of new event timers 
that are not generating interrupts when not needed. In this patch I 
haven't even reimplemented polling support while rewriting that part of 
the code, as for several years since implementing it I haven't seen 
reports that it was really useful. If it is useful, I'll think of it.

-- 
Alexander Motin