Re: [PATCH RFC] virtio-spec: flexible configuration layout

[Sasha Levin <levinsasha928@gmail.com>]

On Wed, 2011-11-09 at 10:47 +0000, Pawel Moll wrote:
> On Wed, 2011-11-09 at 10:20 +0000, Sasha Levin wrote:
> > > > I'm also wondering it it's ok to move virtio configuration out of virtio
> > > > space and into PCI space for archs that don't have PCI (such as ARM).
> 
> Just a note - ARM-based chips can by all means have PCI (grep -r PCI
> arch/arm/ ;-). The fact is that most of the SOCs available on the market
> don't have it, but this is slowly changing.
> 
> The main architectural difference is that ARM doesn't provide separate
> I/O space so the PCI I/O space is usually remapped somewhere into normal
> address space (grep -r "#define __io_address" arch/arm/)
>  
> > > > Would it mean they get stuck with legacy configuration (and no new
> > > > features)? Or is there an alternative for them?
> > > 
> > > The change only affects the layout of virtio PCI. Arches that don't
> > > have PCI don't use virtio PCI, presumably?
> > > 
> > > BTW, the spec only covers x86 ATM, this needs to be fixed.
> > 
> > From what I see there is a WIP by Pawel Moll <pawel.moll@arm.com> to add
> > virtio platform drivers which get virtio working on ARM for example, and
> > by Peter Maydell <peter.maydell@linaro.org> to modify the spec to
> > support MMIO access (besides PCI).
> 
> Yep, it's actually already in 3.2-rc1 (drivers/virtio/virtio_mmio.c) and
> in the spec (see Appendix X). And actually the control registers layout
> I used was originally based on the PCI "legacy" headers (slightly
> simplified), but evolved a bit since. My understanding is that the
> changes Michael is proposing affect the PCI device interface only so
> they shouldn't affect "my" interface.

I didn't know it's in already, might be interesting adding support to it
to x86 userspace tools.

I thought you used the 'legacy' layout, which is why I was worried that
these changed might cause problems for you - but from what I see you
have a different layout there, so as you said, it shouldn't cause any
issues there.

> 
> By the way, I vaguely remember Peter mentioning that he got the PCI
> device "experimentally" running some time ago on one of the PCI-enabled
> ARM platform models (realview or versatile)...


-- 

Sasha.