[Intel-gfx] HDMI Audio Channels mis-detected on i915

[Wu Fengguang]

Hi Jeremy,

On Fri, May 20, 2011 at 08:06:20PM -0500, Jeremy Bush wrote:
> Hello,
> 
> I posted something similar to this on the ALSA mailing list, but they directed me here, since HDMI detection has to do with the video driver instead of ALSA.

Right.

> I have my machine hooked up to an Onkyo TX-SR607 7.2-Channel receiver. Yet when the machine boots,  HDMI detection mis-detects the hardware on the other end of the HDMI cable:
> 
> [ 2.090473] HDMI: detected monitor at connection type HDMI
> [ 2.090475] HDMI: available speakers: FL/FR
> [ 2.090477] HDMI: supports coding type LPCM: channels = 2, rates = 44100 48000 88200, bits = 16
> [ 2.090480] HDMI hot plug event: Pin=6 Presence_Detect=1 ELD_Valid=1

That information is filled by BIOS or VBIOS. There are 3 possibilities
1) BIOS provides some dumb EDID-Like Data (ELD) regardless of the HDMI sink
2) BIOS does not parse EDID correctly
3) your TX-SR607 EDID data is broken

I'm writing a drm patch to do ELD for audio in the graphics driver,
which will address problems caused by (1) and (2). I've done the
scratch patch, but will need more tests before sending it out.

For now you may still manually correct ELD with this command:

        echo speakers 0 > '/proc/asound/card0/eld#3.0'

By clearing the speakers setting, the audio driver will assume your
HDMI sink supports all speakers.

> You can see my whole alsa info (including it actually detecting the hardware can do 7.1 [note Node 0x06 [Pin Complex] wcaps 0x40778d: 8-Channels Digital Amp-Out CP]) here: http://www.alsa-project.org/db/?f=5b55a63fea6580dfcbf1a85c7c734f977df798f1

This is the Intel HDMI port's audio capability and is always
8-channels. In contrast the ELD describes how many speakers (and hence
channels) are attached to the HDMI sink device (TX-SR607 in your case).

> Is this a known problem?
> 
> My machine specs are as follows:
> 
> Intel i3 2100T
> MSI H67MA-E45 (with Realtek ALC892 audio chip)
> 
> I'm using HDMI output only.

Will you try the above workaround command first? For the long term
solution, if you are ready to compile new kernels, I can send you the
patch for testing.

Thanks,
Fengguang