Re: [PATCH 00 of 16] [RFC] amd iommu: support ATS device passthru on IOMMUv2 systems

[Wei Wang2 <wei.wang2@amd.com>]

On Thursday 15 December 2011 11:23:58 Jan Beulich wrote:
> >>> On 14.12.11 at 16:29, Wei Wang <wei.wang2@amd.com> wrote:
> >
> > ATS devices with PRI and PASID capabilities can communicate with iommuv2
> > to do 2 level (nested) DMA translation and IO demand paging. To do that,
> > both iommu driver and ats device have to been enabled in guest OS. This
> > patch set
> >
> > adds initial iommu emulation for hvm guests to support ATS device
> > passthru.
>
> I could take care of the first 6 patches in this series, as they're only
> touching AMD IOMMU code and look sensible to me. I'm not sure
> though whether this is a good idea without knowing the disposition
> of the other 10 patches (particularly the relative large 3rd patch
> doesn't seem to make sense without it later getting hooked up).
>
> Please let me know,
> Jan
 
Actually, the amd specific patches implement the most IOMMUv2 support. Thanks 
for looking at this. We had thought about how to integrate IOMMUv2 for ATS 
device passthru. Since guest OS requires iommu to be presented in this case, 
we could go for either PV interface or full emulation.  The iommuv2 driver 
has just been submitted to Linux mailing list, so this might be too early for 
pv iommu drivers... Using mmio handler in xen, we can avoid any guest OS 
changes and can get better performance than using qemu-dm. So this might be 
the only approach we intend to use at the moment.

But yes, eventually, the iommu emulation is driven by hypercalls. It would be 
great that tools maintainers could check this in the meantime. so, Ian, could 
I invite you to take a look at this?

Many thanks,
Wei