Re: [PATCH 0/3] clean up interface between KVM and psp

[Brijesh Singh]

On 8/17/21 1:08 PM, Mingwei Zhang wrote:
> Hi Paolo,
>
> Thanks for the prompt reply. I will update the code and will be
> waiting for Tom and other AMD folks' feedback.
>
> Thanks. Regards
> -Mingwei
>
> On Tue, Aug 17, 2021 at 1:54 AM Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote:
> > On 16/08/21 22:24, Mingwei Zhang wrote:
> > > This patch set is trying to help make the interface between KVM and psp
> > > cleaner and simpler. In particular, the patches do the following
> > > improvements:
> > >    - avoid the requirement of psp data structures for some psp APIs.
> > >    - hide error handling within psp API, eg., using sev_decommission.
> > >    - hide the serialization requirement between DF_FLUSH and DEACTIVATE.
> > >
> > > Mingwei Zhang (3):
> > >     KVM: SVM: move sev_decommission to psp driver
> > >     KVM: SVM: move sev_bind_asid to psp
> > >     KVM: SVM: move sev_unbind_asid and DF_FLUSH logic into psp
> >
> > No objections apart from the build failure on patch 1.  However, it's up
> > to Tom whether they prefer this logic in KVM or the PSP driver.

I have no objection to move those functions in SEV drv.

With build fix

Acked-by: Brijesh Singh <brijesh.singh@xxxxxxx>


Just for the context, SEV API commands are divided in two sets:

1. commands to provision the host (such as PDH_GEN, CSR, CERT_EXPORT, 
CERT_IMPORT ...)
2. commands to manage the guest (such as LAUNCH_START, LAUNCH_UPDATE ...)

I was trying to keep all the guest management commands functions within 
KVM because no other driver needs it. Having said that, we made 
exception for the decommission and activate so we can cleanup the 
firmware resource in non-process context.

thanks