RE: [PATCH 1/5] xen/common: introduce a new framework for save/restore of 'domain' context

[Paul Durrant <xadimgnik@xxxxxxxxx>]

> -----Original Message-----
[snip]
> >>     * The name of the hypercall does not say anything about "PV". So a
> >> contributor could think it would be fine to save/restore new HVM state
> >> in the new generic hypercall. Is it going to be an issue? If so, how do
> >> we prevent it?
> >
> > The commit message says:
> >
> > "Domain context is state held in the hypervisor that does not come under
> > the category of 'HVM state' but is instead 'PV state' that is common
> > between PV guests and enlightened HVM guests (i.e. those that have PV
> > drivers) such as event channel state, grant entry state, etc."
> 
> This does not seem to cover all the cases. For instance, where would you
> save IOREQ servers information?
> 

Ok, I agree that is ambiguous. I'd still call it PV state but of course it does 
only relate to HVM guests.

> >
> > Do you think this should also appear in a comment? Do we really care? 
> > Nothing fundamentally prevents
> the mechanism being used for HVM state, but that may introduce an ordering 
> dependency.
> 
> I don't particularly like the idea to let the contributor decide where
> "HVM context" or as part of the "Domain context".
> 
> This is  going to result to unwanted dependency and possibly
> bikeshedding on where they should be saved.
> 
> My preference would be to mark the existing framework as deprecated and
> force all the new states to be saved as part of the new "Domain Context".
> 

I'm ok with that. Any others have any opinion to the contrary?

> >
> >>     * Are we going to deprecate the existing framework?
> >>
> >
> > There is no intention as yet.
> >
> >> I am not suggesting we should not go with two frameworks, but the
> >> reasons and implications are not clear to me. Hence, why I think the
> >> commit message should be expanded with some rationale.
> >>
> >
> > Ok, I can add a paragraph to try to explain a little more.
> 
> That would be appreciated. Thank you!
> 

I'll mention the deprecation of the HVM context interface there too.

  Paul