[alsa-devel] HDMI hotplug on Skylake when power well is off

[David Henningsson]

On 2015-07-16 14:21, Takashi Iwai wrote:
> On Thu, 16 Jul 2015 13:35:14 +0200,
> Vinod Koul wrote:
>>
>> On Thu, Jul 16, 2015 at 12:06:51PM +0200, Takashi Iwai wrote:
>>> On Thu, 16 Jul 2015 11:34:02 +0200,
>>> David Henningsson wrote:
>>>>
>>>>
>>>>
>>>> On 2015-07-16 10:39, Vinod Koul wrote:
>>>>> On Wed, Jul 15, 2015 at 12:59:02PM +0200, Takashi Iwai wrote:
>>>>>> On Wed, 15 Jul 2015 11:05:22 +0200,
>>>>>> David Henningsson wrote:
>>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> I'm trying to debug an issue here where the HDMI hotplug events are not
>>>>>>> delivered to the audio side when the power well is off. This is on a
>>>>>>> Skylake machine (running in HDA mode).
>>>>>>>
>>>>>>> I'm not sure whether the problem is upstream or due to my own patches
>>>>>>> while testing, so I was wondering how this is supposed to be working, so
>>>>>>> I can troubleshoot further?
>>>>>>>
>>>>>>> Should there be an IRQ on the HDA controller even if the power well is
>>>>>>> off, and if not, how should the audio driver be notified that an HDMI
>>>>>>> hotplug event has happened?
>>>>>>
>>>>>> I thought this has been always a problem when the runtime PM is
>>>>>> enabled, no matter whether the power well state is.
>>>>> Shouldn't the hotplug action turn on the power well? Then notification for
>>>>> audio side should get propagated as power well is On
>>>>
>>>> While the video side can turn the power well on, maybe there are other
>>>> things that needs to be turned on from the audio driver?
>>>
>>> Yes.  This was supposed to work with WAKEEN bits, but this didn't work
>>> as expected, after all.

But this applies only to HDMI codecs and the power well, right? Analog 
controllers could go into D3 and WAKEEN would wake them up? Or did we 
never get that working either?

Anyhow, I went searching in the video drivers for things to match on, 
inspired by your thoughts about submitting information from the graphics 
driver to the audio driver. See new attached patch (still writing blind 
drafts). Let me know what you think.

I think "connector->name" seems what the most reasonable thing to export 
to userspace to indicate what gfx output is related to which audio nid, 
and the port being what maps 1-to-1 to a nid - we'll use that for our 
own lookup (this lookup might differ between hardware generations).

-- 
David Henningsson, Canonical Ltd.
https://launchpad.net/~diwic
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hdmi_hotplug_notify2.patch
Type: text/x-diff
Size: 7527 bytes
Desc: not available
URL: <http://mailman.alsa-project.org/pipermail/alsa-devel/attachments/20150716/3c8a9456/attachment-0001.bin>