Re: AMD 8131 and MSI quirk

[Kristen Accardi <kristen.c.accardi@intel.com>]

On Wed, 2006-02-15 at 00:10 +0200, Michael S. Tsirkin wrote:
> Quoting r. Kristen Accardi <kristen.c.accardi@intel.com>:
> > I guess the assumption I made was that if msi is
> > turned off for a bridge, then all devices under the bridge may not use
> > msi.   
> 
> Well, all this is just quirks, so no real rules apply, thats why I thought
> having 2 bits gives us maximum flexibility.
> 
> Specifically for PCXH I see this in code:
> /*
>  * It's possible for the MSI to get corrupted if shpc and acpi
>  * are used together on certain PXH-based systems.
>  * */
> 
> So it seems the issue is device-specific - only affects the bridge itself.
> 
> What the code currently does is disable msi for bridge itself but not for
> the devices behind it. I cant inherit dev->no_msi from parent to child
> without changing this, and I just assumed this is by design.
> 
> Are you saying this is a bug and should be changed?
> 

Ok, I went back and read the errata that this patch was attempting to
solve, and you are right, it would not be correct to have children
inherit the no_msi flag from the parent in this case.  I can clearly see
why we need to have a flag associated with the bus rather than the
device for your case.

Kristen