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;
Date:   Mon, 4 Oct 2021 15:14:45 +0200
From:   Frederic Weisbecker <frederic@kernel.org>
To:     Valentin Schneider <valentin.schneider@arm.com>
Cc:     "Paul E . McKenney" <paulmck@kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
        Peter Zijlstra <peterz@infradead.org>,
        Uladzislau Rezki <urezki@gmail.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Boqun Feng <boqun.feng@gmail.com>,
        Neeraj Upadhyay <neeraju@codeaurora.org>,
        Josh Triplett <josh@joshtriplett.org>,
        Joel Fernandes <joel@joelfernandes.org>, rcu@vger.kernel.org
Subject: Re: [PATCH 05/11] rcu/nocb: Make rcu_core() callbacks acceleration
 (de-)offloading safe
Message-ID: <20211004131445.GA273854@lothringen>
References: <20210929221012.228270-1-frederic@kernel.org>
 <20210929221012.228270-6-frederic@kernel.org>
 <87a6jsmy6o.mognet@arm.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <87a6jsmy6o.mognet@arm.com>
Precedence: bulk

On Fri, Oct 01, 2021 at 06:50:39PM +0100, Valentin Schneider wrote:
> On 30/09/21 00:10, Frederic Weisbecker wrote:
> > When callbacks are offloaded, the NOCB kthreads handle the callbacks
> > progression on behalf of rcu_core().
> >
> > However during the (de-)offloading process, the kthread may not be
> > entirely up to the task. As a result some callbacks grace period
> > sequence number may remain stale for a while because rcu_core() won't
> > take care of them either.
> >
> 
> But that should be taken care of at the tail end of the (de)offloading
> process, either by rcu_core() or by the NOCB kthreads, no?

True but the (de-)offloading process can take a random amount of time to
complete. During this time if the queue of callbacks is already huge, things
can explode.

> 
> Or is it e.g. in the case of offloading, we want to make sure an rcu_core()
> invocation runs callback acceleration because even if the NOCB GP/CB
> kthreads are being set up, we're not guaranteed is going to do that
> straight away?

Right.

> 
> IIUC it would be a similar case for deoffload when we stash the NOCB GP/CB
> kthreads and get rcu_core() running concurrently.

You got it!

Thanks.