Re: [PATCH for-4.16 v3] gnttab: allow setting max version per-domain

[Ian Jackson <iwj@xxxxxxxxxxxxxx>]

Roger Pau Monné writes ("Re: [PATCH for-4.16 v3] gnttab: allow setting max 
version per-domain"):
> Would you like me to this to the commit message:
> 
> "xenstored stubdomains are limited to grant table v1 because the
> current MiniOS code used to build them only has support for grants v1.
> There are existing limits set for xenstored stubdomains at creation
> time that already match the defaults in MiniOS."

Yes, please.

> > I am not sure about the implications for migration.  Might using this
> > cause migration to fail for some guests ?
> > 
> >    Note that when using the default grant version the specific max
> >    version in use by the domain is not migrated. Any guests should be
> >    able to cope with the max grant version changing across migrations,
> >    and if a specific guest relies on a maximum grant version being
> >    unconditionally available it should be specified on the configuration
> >    file.
> > 
> > Only if the feature is *not* used, I guess.  Ie, this is the status
> > quo.   So I don't think there is any release risk there.
> 
> This was raised by Jan in a previous version, the discussion can be
> found here:
> 
> https://lore.kernel.org/xen-devel/0b58667f-b6bc-d5b5-2dd1-0c8996367319@xxxxxxxx/
> 
> The issue could arise if a guest that strictly needs grant v2 is
> migrated from a host that has v2 as the default max version to another
> box that has v1 as the max version. If the guest config file doesn't
> explicitly specify that the guest requires grant v2 migration will
> succeed, but the guest will likely fail to resume properly.
> 
> This is already the current behavior if a guest is migrated from a
> host not having gnttab=max-ver set to one having gnttab=max-ver:1.

Right.  I don't think this is the correct behaviour but the patch is
moving in the right direction.

> > If we can get good answers to all of this, ideally I would like to see
> > this committed by the end of tomorrow.  I plan to cut RC1 on Monday.

Thanks.


Tools ack: with the commit message change discussed above,

Reviewed-by: Ian Jackson <iwj@xxxxxxxxxxxxxx>


I have no further release concerns but a formal release-ack will have
to wait for review of the ocaml parts.

Ian.