On Fri, 13 Oct 2023, Julien Grall wrote:
> Hi George,
> 
> On 13/10/2023 17:06, George Dunlap wrote:
> > On Fri, Oct 13, 2023 at 4:56 PM Julien Grall <jul...@xen.org> wrote:
> > > > The initial patch that
> > > > introduced it didn't mention wanting to include dom0 specifically; on
> > > > the contrary it said:
> > > > 
> > > > "The extra notification is harmless for domains that don't require it."
> > > 
> > > ... even if this is what the commit message says, we can't tell whether
> > > someone would start to rely on it. I definitely see use-cases for it.
> > > 
> > > But I agree that the chance they are actualy used is slim as we didn't
> > > document it.
> > 
> > If a dom0 kernel started relying on the behavior, it would be a bug,
> > since it would be incompatible with releases before 4.16.  So I think
> > the only risk here would be if some bit of a system's userspace
> > tooling relied on it; but I can't really see how that would come to
> > be.
> > 
> > If someone reports a bug, we can always add the reporting back, when
> > we're not under time pressure.
> > 
> > > Your reasoning makes sense. I don't fully agree with it, but at the end
> > > of the day we need to have a solution for dom0less guests... So if the
> > > others agree with you, then I will disagree and commit.
> > > 
> > > That said, I think the commit message should explain why removing dom0
> > > notification is ok.
> > 
> > Yes, absolutely.
> > 
> > Do you or Stefano want to pick it up?  I think you'd be able to test
> > it more thoroughly.  I can build-test it and send it to gitlab-CI, but
> > anything more will take a lot more setup.
> 
> I just asked you a similar question on Matrix. :) Stefano was able to
> reproduce the issue. So it would probably be best for him to pick up at least
> the testing part.
> 
> Stefano, do you have some cycle to send the patch as well? If not, I can do
> it.

I have done the work and the patch works:

https://marc.info/?l=xen-devel&m=169723822001933

However, you should know that I tried to reproduce the original race
condition without the patch, but I couldn't.

Reply via email to