From: Alexandre Ratchov <alex@caoua.org>
Subject: Re: azalia(4) testers wanted!
To: Martijn van Duren <openbsd+tech@list.imperialat.at>
Cc: Mark Kettenis <mark.kettenis@xs4all.nl>, tech@openbsd.org, brynet@gmail.com, thfr@openbsd.org, obsd@bartula.de
Date: Fri, 17 May 2024 11:08:51 +0200

On Fri, May 17, 2024 at 09:58:30AM +0200, Martijn van Duren wrote:
> > 
> Thanks, with this things make sense to me. I've been stressing my
> soundcard with multiple streams for about 12 hours now without any
> problems.
> 
> OK martijn@ FWIW
> 
> Going by the comment and your commit message from r1.196, do we still
> need the following section?
>         /* disable MSI, use INTx instead */
>         if (PCI_VENDOR(sc->pciid) == PCI_VENDOR_INTEL) {
>                 reg = azalia_pci_read(sc->pc, sc->tag, ICH_PCI_MMC);
>                 reg &= ~(ICH_PCI_MMC_ME);
>                 azalia_pci_write(sc->pc, sc->tag, ICH_PCI_MMC, reg);
>         }
>

This is probably not needed anymore. But, IIRC, few devices have a
real hardware bug and require MSI to be disabled. Once the diff is in,
we could revisit these workarounds