Re: [RFC PATCH v3 20/21] KVM: x86/xen: Prevent runstate times from becoming negative

From: Steven Rostedt
Date: Tue Aug 20 2024 - 11:08:02 EST


On Tue, 20 Aug 2024 11:22:31 +0100
David Woodhouse <dwmw2@xxxxxxxxxxxxx> wrote:

> On Thu, 2024-08-15 at 21:39 -0700, Sean Christopherson wrote:
> > > +       vx->last_steal = run_delay;
> > > +
> > > +       /*
> > > +        * If KVM clock time went backwards, stop updating until it
> > > +        * catches up (or the runstates are reset by userspace).
> > > +        */
> >
> > I take it this is a legitimate scenario where userpace sets KVM clock and then
> > the runstates, and KVM needs to lend a hand because userspace can't do those two
> > things atomically?
>
> Indeed. Will update the comment to make that more obvious.
>
> Thanks for the rest of the review on this series. I'll go through in
> detail and update it, hopefully this week.

Hmm, is this related at all to this:

https://lore.kernel.org/all/20240806111157.1336532-1-suleiman@xxxxxxxxxx/

-- Steve