Re: NetBSD and Linux KVM

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Wed, Feb 20, 2013 at 09:27:04AM +0900, Minoura Makoto wrote:
> 
> |> In <20130219161238.GB29357%britannica.bec.de@localhost>
> |>  Joerg Sonnenberger <joerg%britannica.bec.de@localhost> wrote:
> 
> > It doesn't make sense. PCI interrupts have a specified behavior, so this
> > is clearly a bug in the MADT.
> 
> Hmm,
> 
>  - It's actually active-high, if I read qemu sources correctly
>  - All of Ubuntu 12.04, CentOS 6 and Windows 7 (installer
>   DVD) configure it as active-high (we can see this by using
>   Linux kernel debugging tools such as Red Hat crash) and
>   work perfectly
> 
> So I think the MADT is correct. 
> 
> I'm not familiar with ACPI specification.  Do you know
> something else to describe interrupt polarity settings?

PCI interrupts require active low, level triggering. ACPI SCI is
somewhat special and can also be active high. ISA is active hi and edge
triggered (IIRC). Point is that different polarities cannot coexist, so
whatever this ACPI table is doing is wrong. Either the override in the
MADT or the PCI routing entry.

Joerg