Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Message-ID: <c44d6ab34f2f4e4a5d36036cc8b356a3f4f3519b.camel@redhat.com>
Subject: Re: [PATCH RT v2 3/3] rcu: Disable use_softirq on PREEMPT_RT
From:   Scott Wood <swood@redhat.com>
To:     Joel Fernandes <joel@joelfernandes.org>
Cc:     Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
        linux-rt-users@vger.kernel.org, linux-kernel@vger.kernel.org,
        "Paul E . McKenney" <paulmck@linux.ibm.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Peter Zijlstra <peterz@infradead.org>,
        Juri Lelli <juri.lelli@redhat.com>,
        Clark Williams <williams@redhat.com>
Date:   Thu, 22 Aug 2019 14:31:17 -0500
In-Reply-To: <20190822135953.GB29841@google.com>
References: <20190821231906.4224-1-swood@redhat.com>
         <20190821231906.4224-4-swood@redhat.com>
         <20190822135953.GB29841@google.com>
Organization: Red Hat
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.30.5 (3.30.5-1.fc29) 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, 2019-08-22 at 09:59 -0400, Joel Fernandes wrote:
> On Wed, Aug 21, 2019 at 06:19:06PM -0500, Scott Wood wrote:
> > I think the prohibition on use_softirq can be dropped once RT gets the
> > latest RCU code, but the question of what use_softirq should default
> > to on PREEMPT_RT remains.
> 
> Independent of the question of what use_softirq should default to, could
> we
> test RT with latest RCU code now to check if the deadlock goes away?  That
> way, maybe we can find any issues in current RCU that cause scheduler
> deadlocks in the situation you pointed. The reason I am asking is because
> recently additional commits [1] try to prevent deadlock and it'd be nice
> to
> ensure that other conditions are not lingering (I don't think they are but
> it'd be nice to be sure).
> 
> I am happy to do such testing myself if you want, however what does it
> take
> to apply the RT patchset to the latest mainline? Is it an achievable feat?

I did run such a test (cherry picking all RCU patches that aren't already in
RT, plus your RFC patch to rcu_read_unlock_special, rather than applying RT
to current mainline) with rcutorture plus a looping kernel build overnight,
and didn't see any splats with or without use_softirq.

-Scott