Re: [RFC PATCH v1 07/25] hw/xen: Implement core serialize/deserialize methods for xenstore_impl

[David Woodhouse]

On Tue, 2023-03-07 at 16:39 +0000, Paul Durrant wrote:
> On 07/03/2023 16:33, David Woodhouse wrote:
> > On Thu, 2023-03-02 at 15:34 +0000, David Woodhouse wrote:
> > > From: David Woodhouse <dwmw@amazon.co.uk>
> > > 
> > > In fact I think we want to only serialize the contents of the domain's
> > > path in /local/domain/${domid} and leave the rest to be recreated? Will
> > > defer to Paul for that.
> > > 
> > > Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
> > 
> > Paul, your Reviewed-by: on this one is conspicuous in its absence. I
> > mentioned migration in the cover letter — this much is working fine,
> > but it's the PV back ends that don't yet work.
> > 
> > I'd quite like to merge the basic serialization/deserialization of
> > XenStore itself, with the unit tests.
> 
> The patch is basically ok, I just think the serialization should be 
> limited to the guest nodes... filtering out those not owned by xen_domid 
> would probably work for that.

Yeah, so let's just do this (as part of this patch #7) for now:

--- a/hw/i386/kvm/xen_xenstore.c
+++ b/hw/i386/kvm/xen_xenstore.c
@@ -235,6 +235,7 @@ static int xen_xenstore_post_load(void *opaque, int
ver)
 
 static const VMStateDescription xen_xenstore_vmstate = {
     .name = "xen_xenstore",
+    .unmigratable = 1, /* The PV back ends don't migrate yet */
     .version_id = 1,
     .minimum_version_id = 1,
     .needed = xen_xenstore_is_needed,


It means we can't migrate guests even if they're only using fully
emulated devices... but I think that's a reasonable limitation until we
implement it fully.

smime.p7s
Description: S/MIME cryptographic signature