Re: [PATCH v4 1/4] KVM: Dynamic sized kvm memslots array

[Michael Tokarev]

21.10.2024 17:37, Peter Xu wrote:
> Michael,
>
> On Fri, Oct 18, 2024 at 06:38:53PM +0300, Michael Tokarev wrote:
> > Looking at this from qemu-stable PoV, I'm not 100% sure this change is good
> > for stable-7.2 series, because 7.2 lacks v8.1.0-1571-g5b23186a95
> > "kvm: Return number of free memslots" commit, which was a preparation for
> > for memory devices that consume multiple memslots.
> >
> > I did a backport of this change (currently it is at the tip of staging-7.2
> > branch of https://gitlab.com/mjt0k/qemu.git) - I had to tweak context and
> > also to remove now-unused local variable in kvm-all.c.  It builds and the
> > tests run fine, but I'm not really sure it does what it is intended to do.
> >
> > Should anything else be picked up for 7.2 for all this to work, or should
> > this change not be back-ported to 7.2 ?
> >
> > (for more recent releases, everything looks ok).
>
> I don't remember anything this series logically depends on (besides any
> context-wise change).

Well, 7.2 is a bit old by now, and the commit I already mentioned above is
also quite old, - at the time you started working on this series, this
commit (v8.1.0-1571-g5b23186a95) has been in the tree for a long time
already.  This change might be relevant here or might be not.

> If there's uncertainty / challenge from backporting to some stable branches
> from your POV, we can still keep things simple and skip the series, as it's
> only a perf regression and only happens during live migrations (which can
> enlarge the downtime, for example) but not daily VM use.

For this change alone, I did the backport, I just am not sure it makes sense.

It would be great if you take a look, including the change I mentioned above
(it isn't in 7.2), there: https://gitlab.com/mjt0k/qemu/-/commits/staging-7.2
Or we can just drop it for 7.2 per the above.

Thanks,

/mjt