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:   Mon, 18 Jul 2022 08:21:32 -0700
From:   "Paul E. McKenney" <paulmck@kernel.org>
To:     Zqiang <qiang1.zhang@intel.com>
Cc:     frederic@kernel.org, quic_neeraju@quicinc.com,
        joel@joelfernandes.org, rcu@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] rcu-tasks: Make RCU Tasks Trace checking for
 userspace execution
Message-ID: <20220718152132.GE1790663@paulmck-ThinkPad-P17-Gen-1>
Reply-To: paulmck@kernel.org
References: <20220718001610.263700-1-qiang1.zhang@intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220718001610.263700-1-qiang1.zhang@intel.com>
Precedence: bulk

On Mon, Jul 18, 2022 at 08:16:10AM +0800, Zqiang wrote:
> For RCU tasks trace, the userspace execution is also a valid quiescent
> state, if the task is in userspace, the ->trc_reader_nesting should be
> zero and if the ->trc_reader_special.b.need_qs is not set, set the
> tasks ->trc_reader_special.b.need_qs is TRC_NEED_QS_CHECKED, this cause
> grace-period kthread remove it from holdout list if it remains here.
> 
> This commit add rcu_tasks_trace_qs() to rcu_flavor_sched_clock_irq()
> when the kernel built with no PREEMPT_RCU.
> 
> Signed-off-by: Zqiang <qiang1.zhang@intel.com>

The looks plausible to me, but can you tell me how this avoids the
following sequence of events?

o	CPU 0 takes a scheduling-clock interrupt.  Just before this
	point CPU 0 was running in user context, thus as you say
	should not be in an RCU Tasks quiescent state.

o	CPU 0 enters an RCU Tasks Trace read-side critical section.

o	CPU 1 starts a new RCU Tasks Trace grace period.

o	CPU 0 reaches the newly added rcu_note_voluntary_context_switch().

	Except that the quiescent state implied by userspace execution
	was before the new grace period, and thus does not apply to it.

(Yes, I know, if this is a bug in this patch, the bug already exists
due to the call in rcu_flavor_sched_clock_irq() for !PREEMPT kernels,
but if this change is safe, it should be possible to explain why.)

							Thanx, Paul

> ---
>  v1->v2:
>  Fix build error due to undeclared rcu_tasks_trace_qs(), note in no-PREEMPT_RCU
>  kernel, the RCU Tasks is replaced by RCU, so rcu_note_voluntary_context_switch()
>  only include rcu_tasks_trace_qs().
> 
>  kernel/rcu/tree_plugin.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index 4152816dd29f..5fb0b2dd24fd 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -976,7 +976,7 @@ static void rcu_flavor_sched_clock_irq(int user)
>  		 * neither access nor modify, at least not while the
>  		 * corresponding CPU is online.
>  		 */
> -
> +		rcu_note_voluntary_context_switch(current);
>  		rcu_qs();
>  	}
>  }
> -- 
> 2.25.1
>