Re: [Xen-devel] [PATCH V2] Add flag to start info regarding virtual mapped p2m list

[Ian Campbell <ian.campbell@xxxxxxxxxx>]

On Tue, 2015-03-17 at 06:50 +0100, Juergen Gross wrote:
> On 03/04/2015 12:41 PM, Ian Campbell wrote:
> > On Wed, 2015-03-04 at 12:22 +0100, Juergen Gross wrote:
> >> It would either be used like intended,
> >
> > Which is how? That is what is really missing here.
> >
> > So far this appears to be a bit which enables some as yet unspecified[0]
> > behaviour in one particular OS kernel with some as yet undiscussed
> > potential future impact on toolstack functionality.
> 
> Sorry for the late answer, but I managed to move this mail to my archive
> before reacting on it. :-(

No problem.

The description looks like the sort of thing which ought to go into the
commit log.

Is there not an ABI change somewhere else relating to the exposure of
the cr3+vaddr+size? If so why is it not in this patch?

Ideally whichever file which needs to change in xen/include/public to
expose that change should also come along with documentation for this
new ABI.

If that change has been deferred for some reason then I think it (and
why) should be mentioned in the commit message, you'll also want to
explain why adding the bit now but the ABI change later is safe, i.e.
what the transition plan is.

AIUI this change has broken memory hotplug and has also made it
difficult from an ABI PoV to reinstate that support. I think that needs
to be addressed (i.e. at the ABI design level, not necessary
implemented) before we add a bit exposing this feature.

> Guests implementing this new interface need to know, of course, whether
> the Xen tools are capable to use the new interface instead of the old
> p2m tree interface. Otherwise a guest using only the new interface with
> the virtual mapped linear p2m list on a machine with old Xen tools not
> supporting this interface could not be restored or migrated.

What is the mechanism for the converse, i.e. for the tools to detect if
a given guest is making use of this functionality?

Are tools going to be able to tell at migration time, or is it something
which needs to be detectable at domain build time (and therefore exposed
in the kernel ELF notes)?

Ian.


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