Re: [Xen-devel] [PATCH V3 2/29] VIOMMU: Add vIOMMU helper functions to create, destroy vIOMMU instance

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Mon, Oct 30, 2017 at 09:41:23AM +0800, Lan Tianyu wrote:
> On 2017年10月25日 09:43, Lan Tianyu wrote:
> >> For all platforms supporting HVM, for PV I don't think it makes sense.
> >> > Since AFAIK ARM guest type is also HVM I would rather introduce this
> >> > field in the hvm_domain structure rather than the generic domain
> >> > structure.
> >> > 
> > This sounds reasonable.
> > 
> >> > You might want to wait for feedback from others regarding this issue.
> >> > 
> > I discussed with Julien before. He hoped no to add viommu code for ARM
> > first.So struct hvm_domain seems to be better place since it's arch
> > specific definition and only add struct viommu for struct hvm_domain of x86.
> 
> Hi Roger:
>       If PV guest needs PV IOMMU support, struct iommu should be put  into
> struct domain and it can be reused by full-virtualization and PV iommu.
> Malcolm Crossley sent out RFC patch of pv iommu before. I found it also
> needs to change struct domain.
> 
> https://lists.xenproject.org/archives/html/xen-devel/2016-02/msg01441.html

This patch series is from February 2016: almost two years old and
there's been no further repost.

If this can indeed be shared with a future pv-iommu work, have you
checked whether the current structure data and hooks would be
suitable for a pv-iommu implementation?

I would rather prefer to move the viommu structure from hvm_domain to
the generic domain struct when it's actually needed (ie: when pv-iommu
is implemented) rather than doing it here.

Roger.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel