Comment 34 for bug 2011385

I've looked further into this and dumped the CORB buffer from booting a windows VM with working sound, using vfio passthrough based on https://github.com/Conmanx360/QemuHDADump

Unfortunately converting these to hda-verb commands and applying on linux didn't fix the internal speaker output. Perhaps I need a particular subset of verbs, but I don't know how to determine which ones.

There are also some pin settings in the CORB frames though, which makes me wonder if patching the pins (e.g. via hdajackretask) is really the problem we have here. Needs further investigation.