Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <CA+G9fYsHo-9tmxCKGticDowF8e3d1RkcLamapOgMQqeP6OdEEg@mail.gmail.com>
 <CANpmjNPpOym1eHYQBK4TyGgsDA=WujRJeR3aMpZPa6Y7ahtgKA@mail.gmail.com>
 <87wnxw86bv.fsf@nanos.tec.linutronix.de> <87eek395oe.fsf@nanos.tec.linutronix.de>
In-Reply-To: <87eek395oe.fsf@nanos.tec.linutronix.de>
From:   Marco Elver <elver@google.com>
Date:   Mon, 7 Dec 2020 13:23:25 +0100
Message-ID: <CANpmjNNdothEQfz6LFN_HHYFQPa6679+WoodMBvsZiPSLndEdw@mail.gmail.com>
Subject: Re: BUG: KCSAN: data-race in tick_nohz_next_event / tick_nohz_stop_tick
To:     Thomas Gleixner <tglx@linutronix.de>
Cc:     Naresh Kamboju <naresh.kamboju@linaro.org>,
        open list <linux-kernel@vger.kernel.org>,
        kasan-dev <kasan-dev@googlegroups.com>, rcu@vger.kernel.org,
        lkft-triage@lists.linaro.org,
        Peter Zijlstra <peterz@infradead.org>,
        "Paul E. McKenney" <paulmck@kernel.org>,
        Ingo Molnar <mingo@kernel.org>, fweisbec@gmail.com,
        Arnd Bergmann <arnd@arndb.de>,
        Dmitry Vyukov <dvyukov@google.com>,
        linux-toolchains@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Sun, 6 Dec 2020 at 00:47, Thomas Gleixner <tglx@linutronix.de> wrote:
> On Sat, Dec 05 2020 at 19:18, Thomas Gleixner wrote:
> > On Fri, Dec 04 2020 at 20:53, Marco Elver wrote:
> > It might be useful to find the actual variable, data member or whatever
> > which is involved in the various reports and if there is a match then
> > the reports could be aggregated. The 3 patterns here are not even the
> > complete possible picture.
> >
> > So if you sum them up: 58 + 148 + 205 instances then their weight
> > becomes more significant as well.
>
> I just looked into the moderation queue and picked stuff which I'm
> familiar with from the subject line.

We managed to push (almost) everything that was still in private
moderation to public moderation, so now there's even more to look at:
https://syzkaller.appspot.com/upstream?manager=ci2-upstream-kcsan-gce
:-)

> There are quite some reports which have a different trigger scenario,
> but are all related to the same issue.
>
>   https://syzkaller.appspot.com/bug?id=f5a5ed5b2b6c3e92bc1a9dadc934c44ee3ba4ec5
>   https://syzkaller.appspot.com/bug?id=36fc4ad4cac8b8fc8a40713f38818488faa9e9f4
>
> are just variations of the same problem timer_base->running_timer being
> set to NULL without holding the base lock. Safe, but insanely hard to
> explain why :)
>
> Next:
>
>   https://syzkaller.appspot.com/bug?id=e613fc2458de1c8a544738baf46286a99e8e7460
>   https://syzkaller.appspot.com/bug?id=55bc81ed3b2f620f64fa6209000f40ace4469bc0
>   https://syzkaller.appspot.com/bug?id=972894de81731fc8f62b8220e7cd5153d3e0d383
>   .....
>
> That's just the ones which caught my eye and all are related to
> task->flags usage. There are tons more judging from the subject
> lines.
>
> So you really want to look at them as classes of problems and not as
> individual scenarios.

Regarding auto-dedup: as you suggest, it'd make this straightforward
if we had the variable name -- it turns out that's not so trivial. I
think we need compiler support for that, or is there some existing
infrastructure that can just tell us the canonical variable name if it
points into a struct or global? For globals it's fine, but for
arbitrary pointers that point into structs, I don't see how we could
do it without compiler support e.g. mapping PC->variable name (we need
to map instructions back to the variable names they access).

Any precedence for this? [+Cc linux-toolchains@vger.kernel.org]

Thanks,
-- Marco