Hi,

On Wed, 19 Jul 2023, Maarten Lankhorst wrote:

On Tue, 18 Jul 2023 19:04:41 +0200, Kai Vehmanen wrote:

...

My only bigger concern is corner cases where the display PCI device is

on

...

the bus and visible to kernel, but for some reason there is no working driver in the system or it is disabled.

Yeah, I have no answer for this. My guess is that in an ideal world, the optional features related to HDMI outputs would be put in a separate sub-driver, which could -EPROBE_DEFER. Only when this driver loads, features related to display will work, but the main audio driver could still load.

in longer term, we have ongoing work in SOF to allow exposing multiple cards (e.g. to have a separate card for HDMI/DP PCM devices), and we are continuously working at improving the data we get from ACPI to have less guesswork in the driver. But this really doesn't help in the shortterm and/or cover all scenarios.

So for now, this is legacy we just need to deal with. OTOH, I do agree that...

A module option to snd_hdac_i915_init would probably be the least of all evils here.

I see the removal of the 60 second timeout as a good thing regardless. :-) Usually when nomodeset is used, it's just for safe mode.

With the addition of  the xe driver, blindly modprobing i915 will fall apart regardless.

The modprobing of i915 from the audio driver, has always felt a bit out-of-place, and with the xe driver, this simply won't scale anymore.

The test results so far look good and this patchset works ok even if some of the more complex multi-GPU configurations we have, so I think with a module option to snd_hdac_i915, I'd be ready to go with this.

Br, Kai