[libvirt] [PATCH 3/8] qemu: Fix memory locking limit calculation

[Andrea Bolognani]

On Mon, 2017-03-27 at 15:42 +0200, Martin Kletzander wrote:
> > > Are you planning on adding new element to the
> > > domain XML which would allow setting this number as well?
> > 
> > I do. Unless I forget about it again, of course :)
> 
> Well, honestly, then I feel really bad about forcing people do different
> choices and changing it between releases.  I don't think anyone wants to
> be checking all documentation changes every release.  But since there
> are only two releases with the patch being in, it's probably okay.  But
> I would *at least* be nice to mention that the lock limit is on it's way
> to the XML.

I'm not going to advertise a feature before it has
materialized: this is not Hollywoo, we don't need to get
people hyped using teasers and trailers ;)

Moreover, I'm actually starting to question the utility of
such a feature altogether.

If you're using <memoryBacking><locked>, all the memory
allocated by QEMU is going to be locked, so there's no
point in setting the memory locking limit to anything but
<memtune><hard_limit>.

For all other cases our calculations should be correct, and
if it turns out that they aren't we want people to get back
to us and let us know so we can fix it for everybody rather
than work around the bug using the new setting.

-- 
Andrea Bolognani / Red Hat / Virtualization