On Fri, Dec 30, 2022 at 1:02 AM Joel Fernandes <joel@xxxxxxxxxxxxxxxxx> wrote:
>
> On Thu, Dec 29, 2022 at 4:10 PM Joel Fernandes <joel@xxxxxxxxxxxxxxxxx> wrote:
> >
> >
> >
> > > On Dec 28, 2022, at 9:39 PM, Joel Fernandes <joel@xxxxxxxxxxxxxxxxx> wrote:
> > >
> > > Hello,
> > > Please apply to v5.10 stable: 4d60b475f858 ("rcu: Prevent lockdep-RCU splats
> > > on lock acquisition/release"). The patch made it in v5.11
> > >
> > > Without it, I get the follow splat on TREE05 rcutorture testing:
> > >
> > > [ 1.253678] CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.10.162-rc1+ #6
> > > [ 1.253678] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.15.0-1 04/01/2014
> > > [ 1.253678] Call Trace:
> > > [ 1.253678]
> > > [ 1.253678] =============================
> > > [ 1.253678] WARNING: suspicious RCU usage
> > > [ 1.253678] 5.10.162-rc1+ #6 Not tainted
> > > [ 1.253678] -----------------------------
> > > [ 1.253678] kernel/kprobes.c:300 RCU-list traversed in non-reader section!!
> > >
> > > I tested with the patch and the warning is gone.
> >
> > Please hold off on this particular one. After a 12 hour test, I am seeing some new warnings. Will update the thread after some more debug.
> >
>
> Weirdly enough, the warning seems to be a false-positive from the
> rcutorture scripts:
>
> WARNING: Assertion failure in
> /root/.jenkins/workspace/rcutorture_stable_linux-5.10.y@2/tools/testing/selftests/rcutorture/res/2022.12.29-02.23.56/TREE05/console.log
> TREE05
> WARNING: Summary:
>
> But opening the console.log shows a normal flow followed by SUCCESS
> message from the rcutorture kernel module.
> Also "Summary:" is supposed to provide more information.
>
> Here are some logs:
> 1. http://box.joelfernandes.org:9080/job/rcutorture_stable/job/linux-5.10.y/10/console
> (Scroll to the end)
> 2. http://box.joelfernandes.org:9080/job/rcutorture_stable/job/linux-5.10.y/10/artifact/tools/testing/selftests/rcutorture/res/2022.12.29-02.23.56/RUDE01/console.log
>
> So this is probably something wrong with my setup, say something
> interfered with the test. Or, there is something weird with the
> scripts.
>
> Anyway, re-running it with this patch overnight again!
I think I found it, v5.10 is also missing this patch which made it in
v5.11: 8d68e68a781d ("torture: Exclude "NOHZ tick-stop error" from
fatal errors")
That is needed along with the $SUBJECT patch.
I will re-run my 5.10 tests again with both these to see if everything
is now green.
- Joel
