Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Thu, 29 Nov 2018 12:23:21 +0100
From:   Heiko Carstens <heiko.carstens@de.ibm.com>
To:     Thomas Gleixner <tglx@linutronix.de>
Cc:     Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@kernel.org>,
        Martin Schwidefsky <schwidefsky@de.ibm.com>,
        linux-kernel@vger.kernel.org, linux-s390@vger.kernel.org,
        Stefan Liebler <stli@linux.ibm.com>
Subject: Re: WARN_ON_ONCE(!new_owner) within wake_futex_pi() triggered
References: <20181127081115.GB3625@osiris>
 <alpine.DEB.2.21.1811281526430.1532@nanos.tec.linutronix.de>
In-Reply-To: <alpine.DEB.2.21.1811281526430.1532@nanos.tec.linutronix.de>
Message-Id: <20181129112321.GB3449@osiris>
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: 8BIT
MIME-Version: 1.0
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, Nov 28, 2018 at 03:32:45PM +0100, Thomas Gleixner wrote:
> Heiko,
> 
> On Tue, 27 Nov 2018, Heiko Carstens wrote:
> 
> > with the glibc self-tests I was able to trigger the "this should not
> > happen" warning ;) below on s390 (with panic_on_warn=1 set). It looks
> > like it is hardly reproducible.
> 
> Any idea which self-test triggered that?
> 
> > This one happened with commit d146194f31c9 for compiling the kernel.
> > Config can be re-created with "make ARCH=s390 performance_defconfig".
> 
> Which is not really helpful for people who do not own a s390. And no, I
> don't want one unless IBM pays the power bill as well :)
> 
> > [  649.596938] WARNING: CPU: 0 PID: 58886 at kernel/futex.c:1418 do_futex+0xa9a/0xc50
> > [  649.596946] Kernel panic - not syncing: panic_on_warn set ...
> > [  649.596951] CPU: 0 PID: 58886 Comm: ld64.so.1 Not tainted 4.20.0-20181125.rc3.git0.d146194f31c9.300.fc29.s390x+git #1
> 
> That's ld64.so.1. Weird, but what do I know about glibc self tests.
> 
> I still fail to see how that can happen, but I usually page out the futex
> horrors immediately. I'll keep staring at the code...

I looked into the system dumps, and if I didn't screw up, then the
command line for both occurrences was

/root/glibc-build/nptl/tst-robustpi8

And indeed, if I run only this test case in an endless loop and do
some parallel work (like kernel compile) it currently seems to be
possible to reproduce the warning:

while true; do time ./testrun.sh nptl/tst-robustpi8 --direct ; done

within the build directory of glibc (2.28).

See
https://sourceware.org/git/?p=glibc.git;a=blob;f=nptl/tst-robustpi8.c;h=cbea3d6d77abb00be05ec7b466d8339c26dd2efb;hb=3c03baca37fdcb52c3881e653ca392bba7a99c2b

which includes this one:

https://sourceware.org/git/?p=glibc.git;a=blob;f=nptl/tst-robust8.c;h=9c636250d4cb0bcd6d802910e8f9ea31568bb73f;hb=3c03baca37fdcb52c3881e653ca392bba7a99c2b