Re: [Xen-devel] [PATCH for 4.6 v2 3/3] xl/libxl: disallow saving a guest with vNUMA configured

[Ian Campbell <ian.campbell@xxxxxxxxxx>]

On Wed, 2015-09-09 at 21:21 +0100, Wei Liu wrote:
> On Wed, Sep 09, 2015 at 07:49:16PM +0100, Andrew Cooper wrote:
> > On 09/09/15 18:03, Wei Liu wrote:
> > > This is due to migration v2 frame record doesn't contain node
> > > information.
> > 
> > This isn't a migration v2 bug, and it was similarly non-functional with
> > legacy migration.  It was yet another oversight with the vNUMA work.
> > 
> 
> Indeed. I mentioned that in v1 patch but dropped it in v2 -- legacy
> migration is gone anyway.

I think it is important in these cases to understand if this is a
functional regression vs 4.5 or not, so it ought to be mentioned.

> > Isn't this information available in the domain configuration for the
> > domain
> > sent in the xl header? (That layer violation also need removing).
> > 
> 
> Yes. But restoring a guest takes a path  different from guest creation
> to populate guest pages.
> 
> > Finally, your phrasing is somewhat unclear.  I would recommend "The
> > migration stream does not preserve node information" as a clearer
> > alternative.
> > 
> 
> Fine by me.

I'm expecting a v3 then.


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