Re: [PATCH RFC 2/2] x86/mm: PG_translate implies PG_refcounts

[Tim Deegan <tim@xen.org>]

At 11:05 +0100 on 24 Aug (1503572736), Wei Liu wrote:
> On Thu, Aug 24, 2017 at 11:02:33AM +0100, Tim Deegan wrote:
> > At 16:58 +0100 on 23 Aug (1503507504), Wei Liu wrote:
> > > After 404595352 ("x86/paging: Enforce PG_external == PG_translate ==
> > > PG_refcounts"), PG_refcounts and PG_translate always need to be set
> > > together.
> > > 
> > > Squash PG_refcounts to simplify code. All calls paging_mode_refcounts
> > > are replaced by calls to paging_mode_translate.
> > 
> > Would it be a good idea to merge all three and have only PG_external?
> > 
> 
> My reverse-engineering is that when PV guest is migrated it has
> PG_external and (the new) PG_translate.
> 
> I would be happy to squash all three into one if you tell me I'm wrong.
> :-)

I _think_ you're wrong :) but can't check right now as I'm
semi-offline.  Migrating PV guests should have paging enabled but
not any of external, translate or refcounts.

There was once code that used PG_translate with PV guests but never
AFAIK checked in to upstream Xen.  There was some talk of using that
combination for PVH at one point but IIRC PVH now uses
translate|refcounts|external like HVM.

Cheers,

Tim.

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