Re: pci-passthrough and shared interrupts

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Wed, Aug 25, 2010 at 03:48:35PM -0700, Ritu kaur wrote:
> Thanks Konrad. Additional Q below
> 
> On Wed, Aug 25, 2010 at 3:34 PM, Konrad Rzeszutek Wilk <
> konrad.wilk@oracle.com> wrote:
> 
> > On Wed, Aug 25, 2010 at 03:17:29PM -0700, Ritu kaur wrote:
> > > Hello,
> > >
> > > Couple of months back when we were testing xcp 0.1.1 and pci passthrough
> > we
> > > ran into issues when pci-passthrough enabled device shared interrupts
> > with
> > > another controller(in our case interrupt was shared between ide
> > controller
> > > and our nic). At that time Ian Campbell suggested backporting
> > event-channel
> > > changes from pvops to 2.6.27 kernel. Have two questions related to this
> > >
> > > 1. I am thinking of porting the changes now but wanted to check if it has
> > > been addressed?
> >
> > I've added some patches in the pvops tree that make this work. You would
> > essentially have to backport the pciback driver to the XCP
> > infrastructure. And you might have to backport the event channels as
> > well, not sure about that.
> >
> 
>  I will take a look at the patches.
> 
> > >
> > > 2. Will shared interrupts be a problem for pci-e slots as well?
> >
> > No.
> >
> 
> So if I have XCP + NIC installed in pci-e slot and interrupts shared with
> another device, pci passthrough should work fine, is this correct?

Ugh, I am surprised you have that. Most of the time if you do use an NIC
in PCI-e it would be MSI based. An MSI's are never shared.

So if your device is PCIe and you get legacy IRQ that is shared with
another device on your board, then you will need to backport the
'pciback' driver.

What NIC is it that does this?