MIME-Version: 1.0
References: <CAHk-=wii6BZtVKYfvQCQqbE3+t1_yAb-ea80-3PcJ4KxgpfHkA@mail.gmail.com>
 <d915df60-d06b-47d4-8b47-8aa1bbc2aac7@roeck-us.net>
In-Reply-To: <d915df60-d06b-47d4-8b47-8aa1bbc2aac7@roeck-us.net>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Mon, 13 Mar 2023 11:21:44 -0700
Message-ID: <CAHk-=wi8xQw6eTRncbJfWWYHHi0UBG2B_AfHrzZbPr=k0V_WYQ@mail.gmail.com>
Subject: Re: Linux 6.3-rc2
To:     Guenter Roeck <linux@roeck-us.net>,
        "Paul E. McKenney" <paulmck@kernel.org>,
        Frederic Weisbecker <frederic@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>
Cc:     Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Mon, Mar 13, 2023 at 8:53=E2=80=AFAM Guenter Roeck <linux@roeck-us.net> =
wrote:
>
> Warning backtraces in calls from ct_nmi_enter(),
> seen randomly.

Hmm.

I suspect this one is a bug in the warning, not in the kernel,
although I have no idea why it would have started happening now.

This happens from an irq event, but that check is not *supposed* to
happen at all from interrupts:

         * We dont accurately track softirq state in e.g.
         * hardirq contexts (such as on 4KSTACKS), so only
         * check if not in hardirq contexts:

but I think that the ct_nmi_enter() function was called before the
hardirq count had even been incremented.

> Sample decoded stack trace:

Hmm. That WARNING backtrace doesn't actually seem to follow the stack
chain, so it only shows the irq stack, not where the irq happened.

> Seen if CONFIG_DEBUG_LOCK_ALLOC=3Dy and CONFIG_CONTEXT_TRACKING_IDLE=3Dy.
> It seems that rcu_read_lock_sched_held() can be true when entering an int=
errupt.
>
> The problem is not seen in v6.2, but occurs randomly on ToT with various
> arm emulations.

Strange. I must be wrong about this being a race on the warning
itself, because that warning has been there for a long long time.

Adding in some people who might have more of a clue. I'm thinking
Frederic and Paul might know what's up with the context tracking, but
I don't see why this would be arm-related or have started recently.
But I do note that PeterZ did some rcuidle tracing cleanups that do
end up affecting arm too.

So adding PeterZ too.

Original email with full details at

   https://lore.kernel.org/lkml/d915df60-d06b-47d4-8b47-8aa1bbc2aac7@roeck-=
us.net/

for added peeps.

Anybody?

                     Linus