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:   Wed, 3 Aug 2022 10:42:06 -0700
From:   "Paul E. McKenney" <paulmck@kernel.org>
To:     Ingo Molnar <mingo@kernel.org>
Cc:     Steven Rostedt <rostedt@goodmis.org>,
        Liu Song <liusong@linux.alibaba.com>, mingo@redhat.com,
        peterz@infradead.org, juri.lelli@redhat.com,
        vincent.guittot@linaro.org, dietmar.eggemann@arm.com,
        bsegall@google.com, mgorman@suse.de, bristot@redhat.com,
        vschneid@redhat.com, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] sched/debug: avoid executing show_state and causing rcu
 stall warning
Message-ID: <20220803174206.GH2125313@paulmck-ThinkPad-P17-Gen-1>
Reply-To: paulmck@kernel.org
References: <1659489525-82994-1-git-send-email-liusong@linux.alibaba.com>
 <20220803084235.5d56d1e4@gandalf.local.home>
 <20220803140653.GD2125313@paulmck-ThinkPad-P17-Gen-1>
 <YuqtDA4NY/8Z22aK@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YuqtDA4NY/8Z22aK@gmail.com>
Precedence: bulk

On Wed, Aug 03, 2022 at 07:14:52PM +0200, Ingo Molnar wrote:
> 
> * Paul E. McKenney <paulmck@kernel.org> wrote:
> 
> > There is an rcu_sysrq_start() and rcu_sysrq_end() to suppress this.  
> > These are invoked by __handle_sysrq().  The value of 
> > rcu_cpu_stall_suppress should be non-zero during the sysrq execution, and 
> > this should prevent RCU CPU stall warnings from being printed.
> > 
> > That said, the code currently does not support overlapping calls to the 
> > various functions that suppress RCU CPU stall warnings.  Except that the 
> > only other use in current mainline is rcu_panic(), which never 
> > unsuppresses.
> > 
> > So could you please check the value of rcu_cpu_stall_suppress? Just in 
> > case some other form of suppression was added somewhere that I missed?
> 
> So instead of supressing the (justified!) RCU stall messages, I'd much 
> rather we apply only the minimal locking necessary for this debug printout.
> 
> That should also solve the stall warnings as a side effect.

I am certainly with you in spirit!  If I recall correctly, the issue
that led to the current state was that there was no way to walk the
task list locklessly except under an RCU read-side critical section.
Yes, you can use get_task_struct(), but that only prevents that task
structure from being freed, not from being removed from the list.

Here is hoping that there is a better way to nail down a task while
RCU-pausing a task-list traversal.  Thoughts?

							Thanx, Paul