Re: [Qemu-devel] [RFC 0/4] virtio-mmio transport

[Pawel Moll]

On Mon, 2011-12-12 at 15:11 +0000, Stefan Hajnoczi wrote:
> If there aren't already then pretty soon ARM-based systems will deal
> with PCIe and Message Signalled Interrupts. 

Actually PCI is not an alien in ARM world - we had platforms with PCI
long time ago. And new SOCs aimed at servers come with PCIe indeed. It's
just that bulk of available systems were designed for mobile phones
etc., and I think you agree that in such case not "wasting" silicon on
huge PCI root complex makes at least some sense...

And generally, when you have PCI-ed platform, you just use
virtio-pci :-)

> Are you sure new ARM
> boards are constraints to a small number of physical interrupts?

It's about SOC rather then a board. If I remember correctly Cortex-A9's
GIC (interrupt controller) can be synthesized with up to 224 interrupt
inputs. And I've already seen SOCs with more interrupt sources in the
past...

Having said all this, I don't think it is really a problem now. At least
today virtio-mmio will be used in simple models rather that in huge
KVM/Xen/whatever virtualised systems with hundreds of PV devices.

Cheers!

Paweł