Re: [RFC] kvm tools: Add support for virtio-mmio

[Sasha Levin <levinsasha928@gmail.com>]

On Tue, 2011-11-15 at 17:00 +0000, Peter Maydell wrote:
> On 15 November 2011 16:47, Sasha Levin <levinsasha928@gmail.com> wrote:
> > +       vmmio->hdr = (struct virtio_mmio_hdr) {
> > +               .magic          = {'v', 'i', 'r', 't'},
> > +               .version        = 1,
> > +               .device_id      = device_id - 0x1000 + 1,
> > +               .vendor_id      = PCI_VENDOR_ID_REDHAT_QUMRANET,
> > +               .queue_num_max  = 256,
> > +       };
> 
> This isn't a PCI device, so does it make sense to use a PCI vendor
> ID here? The kernel doesn't check the vendor ID at the moment,
> but presumably the idea of the field is to allow the kernel to
> work around implementation bugs/blacklist/whatever if necessary.
> If that's the theory then it would make more sense for QEMU and
> kvm-tool to use IDs that say "this is the QEMU implementation"
> and "this is the kvm-tool implementation".
> 
> (I picked 0x554D4551 for QEMU...)
> 
> -- PMM

I just sheepishly filled in the only vendor ID I knew of in the virtio
spec :)
 
Hmm... If thats the plan, it should probably be a virtio thing (not
virtio-mmio specific).

Either way, it could also use some clarification in the spec.

-- 

Sasha.