Re: [PULL 00/16] migration queue

[Christian Schoenebeck]

On Samstag, 31. Oktober 2020 20:10:49 CET Christian Schoenebeck wrote:
> On Samstag, 31. Oktober 2020 18:46:11 CET Peter Xu wrote:
> > On Sat, Oct 31, 2020 at 05:26:28PM +0000, Peter Maydell wrote:
> > > On Sat, 31 Oct 2020 at 16:12, Christian Schoenebeck
> > > 
> > > <qemu_oss@crudebyte.com> wrote:
> > > > On Montag, 26. Oktober 2020 17:19:36 CET Dr. David Alan Gilbert (git)
> 
> wrote:
> > > > > ----------------------------------------------------------------
> > > > > migration pull: 2020-10-26
> > > > > 
> > > > > Another go at Peter's postcopy fixes
> > > > > 
> > > > > Cleanups from Bihong Yu and Peter Maydell.
> > > > > 
> > > > > Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
> > > > 
> > > > May it be possible that this PR introduced a lockup of the qtests that
> > > > I
> > > > am
> > > > encountering in this week's upstream revisions?
> > > 
> > > If you try the patches Peter Xu attached to this thread
> > > does the lockup go away ?
> > > 
> > > https://lore.kernel.org/qemu-devel/20201030135350.GA588069@xz-x1/
> > > 
> > > (I'm also seeing intermittent hangs, for some reason almost always
> > > on s390x host.)
> > 
> > It would be good to know exactly which test hanged.  If it's
> > migration-test
> > then it's very possible.
> 
> It's run-test-144 that does not return; according to Makefile.mtest that's
> migration-test, so chances are high that it's indeed introduced by this PR.
> 
> > The race above patch(es) tried to fix should logically be reproducable on
> > all archs, not s390x only.
> > 
> > Thanks,
> 
> Yes, it's i386 here that locks up.
> 
> I'm running the loop with your patches now, so far so good, let's see if
> it's still alive tomorrow.
> 
> Best regards,
> Christian Schoenebeck

Looks good! 16h later and the loop is still running here; it also made the 
lockup to disappear on Travis-CI. So Peter Xu's two patches fix the lockup 
problem for me.

Best regards,
Christian Schoenebeck