Re: [patch 21/32] NTB/msi: Convert to msi_on_each_desc()

[Dave Jiang]

On 12/1/2021 2:44 PM, Thomas Gleixner wrote:
> On Wed, Dec 01 2021 at 14:21, Dave Jiang wrote:
> > On 12/1/2021 1:25 PM, Thomas Gleixner wrote:
> > > > The hardware implementation does not have enough MSIX vectors for
> > > > guests. There are only 9 MSIX vectors total (8 for queues) and 2048 IMS
> > > > vectors. So if we are to do MSI-X for all of them, then we need to do
> > > > the IMS backed MSIX scheme rather than passthrough IMS to guests.
> > > Confused. Are you talking about passing a full IDXD device to the guest
> > > or about passing a carved out subdevice, aka. queue?
> > I'm talking about carving out a subdevice. I had the impression of you
> > wanting IMS passed through for all variations. But it sounds like for a
> > sub-device, you are ok with the implementation of MSIX backed by IMS?
> I don't see anything wrong with that. A subdevice is it's own entity and
> VFIO can chose the most conveniant representation of it to the guest
> obviously.
>
> How that is backed on the host does not really matter. You can expose
> MSI-X to the guest with a INTx backing as well.
>
> I'm still failing to see the connection between the 9 MSIX vectors and
> the 2048 IMS vectors which I assume that this is the limitation of the
> physical device, right?

I think I was confused with what you were asking and was thinking you 
are saying why can't we just have MSIX on guest backed by the MSIX on 
the physical device and thought there would not be enough vectors to 
service the many guests. I think I understand what your position is now 
with the clarification above.


> What needs a subdevice to expose?
>
> Thanks,
>
>          tglx