Re: [Qemu-devel] [RFC] Device isolation infrastructure v2

[Aaron Fabbri <aafabbri@cisco.com>]

On 12/20/11 8:30 PM, "Alex Williamson" <alex.williamson@redhat.com> wrote:

> On Wed, 2011-12-21 at 14:32 +1100, David Gibson wrote:
>> On Mon, Dec 19, 2011 at 04:41:56PM +0100, Joerg Roedel wrote:
>>> On Mon, Dec 19, 2011 at 11:11:25AM +1100, David Gibson wrote:
<snip>
>>> 
>>> Well, the iommu-api was designed for amd-vi and vt-d. But its concepts
>>> turn out to be more general and by no way x86-centric anymore.
>> 
>> It's improving, but there are still plenty of x86isms there.
> 
> Having worked on ia64 for a while, it's interesting to see this x86
> bashing from the other side.  Everyone is more than willing to make
> architecture neutral interfaces (jeez, look at the extent of the vfio
> reworks), but it's not fair to throw away interfaces as x86-centric if
> you're not pushing your requirements and making use of the code.
> 
> It seems like we'd be better served today to start with the vfio code we
> have and let that be the catalyst to drive an iommu api that better
> serves non-x86.  I don't see how this group management tangent is really
> getting us anywhere.  Thanks,

I'd agree that incremental approach here is key.  VFIO has already seen a
ton of rework to accommodate all architectures.  Let's not bite off a bunch
of these other subsystem rewrites in the same chunk as our VFIO effort.

-Aaron