Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 17 Mar 2022 13:08:15 +0100
From:   Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     Steven Rostedt <rostedt@goodmis.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>
Subject: Re: sched_core_balance() releasing interrupts with pi_lock held
Message-ID: <YjMkr4bE4Gws189a@linutronix.de>
References: <20220308161455.036e9933@gandalf.local.home>
 <20220315174606.02959816@gandalf.local.home>
 <20220316202734.GJ8939@worktop.programming.kicks-ass.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <20220316202734.GJ8939@worktop.programming.kicks-ass.net>
Precedence: bulk

On 2022-03-16 21:27:34 [+0100], Peter Zijlstra wrote:
> Now, we only do queue_core_balance() from set_next_task_idle(), which
> *should* only happen from pick_next_task(), and as such the callback
> should only ever get called from finish_lock_switch() or the 'prev ==
> next' case in __schedule().
> 
> Neither of these two sites holds pi_lock.

I've been trying to reproduce it and didn't make it. I see only the
idle/scheduler path.

> This is about as far as I got explaining things, and it being late, it's
> about as far as I got looking at things.
> 
> Now that also makes conceptual sense, we only want to pull a core-cookie
> task when we're scheduling an idle task.
> 
> Now, clearly this gets triggered from the PI path, but that's not making
> immediate sense to me, it would mean we're boosting the idle task, which
> is wrong too.

Looking at the idle task, it shouldn't be possible for !RT due to lack
of boostable locks and I don't see anything sleeping locks here on RT
either. 

> So it would be useful for someone that can reproduce this to provide a
> trace of where queue_core_balance() gets called, because that *should*
> only be in __schedule().

I failed so far.

Sebastian