Re: [Xen-devel] [PATCH 02/23] PVH xen: turn gdb_frames/gdt_ents into union.

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Mon, 2013-08-05 at 14:51 -0700, Mukesh Rathor wrote:
> On Mon, 5 Aug 2013 16:55:22 +0100
> George Dunlap <George.Dunlap@xxxxxxxxxxxxx> wrote:
> 
> > On Sat, Jul 20, 2013 at 2:44 AM, Mukesh Rathor
> > <mukesh.rathor@xxxxxxxxxx> wrote:
> > > Changes in V2:
> > >   - Add __XEN_INTERFACE_VERSION__
> > >
> > >   Changes in V3:
> > >     - Rename union to 'gdt' and rename field names.
> > >
> > >   Change in V9:
> > >    - Update __XEN_LATEST_INTERFACE_VERSION__ to 0x00040400 for
> > > compat.
> > >
> > > Signed-off-by: Mukesh Rathor <mukesh.rathor@xxxxxxxxxx>
> > > Reviewed-by: Jan Beulich <jbeulich@xxxxxxxx>
> > 
> > One thing that's missing here is a description of *why* this change is
> > being made.  Seeing that it's introducing a union isn't too difficult;
> > harder is figuring out why that's necessary.
> > 
> > Presumably this is to more closely reflect how an HVM guest's GDT is
> > stored -- i.e., in the guest's memory and checked by the hardware on
> > use, rather than in Xen's memory, and checked by Xen on assignment?
> 
> Right. Unlike PV which passes it's gdt pages to xen to be installed, a
> PVH only passes the GDT base and size, as it manages its own GDT. I'll
> add more to the comments.

If the guest is managing its own gdt can't it also use lgdt instructions
etc and expect the hypervisor to pull the GDT info out of the VMCS when
it needs it?

Presumably there is a reason why this doesn't work, perhaps because the
gdt is specified by virtual address? (This ring a bell from a past
conversation). It would be good to elaborate in the commit log.

Ian.


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