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:   Thu, 15 Oct 2020 11:50:33 +0200
From:   Peter Zijlstra <peterz@infradead.org>
To:     "Paul E. McKenney" <paulmck@kernel.org>
Cc:     Boqun Feng <boqun.feng@gmail.com>, Qian Cai <cai@redhat.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Ingo Molnar <mingo@kernel.org>, x86 <x86@kernel.org>,
        linux-kernel@vger.kernel.org, linux-tip-commits@vger.kernel.org,
        Linux Next Mailing List <linux-next@vger.kernel.org>,
        Stephen Rothwell <sfr@canb.auug.org.au>
Subject: Re: [tip: locking/core] lockdep: Fix lockdep recursion
Message-ID: <20201015095033.GS2651@hirez.programming.kicks-ass.net>
References: <20201013112544.GZ2628@hirez.programming.kicks-ass.net>
 <20201013162650.GN3249@paulmck-ThinkPad-P72>
 <20201013193025.GA2424@paulmck-ThinkPad-P72>
 <20201014183405.GA27666@paulmck-ThinkPad-P72>
 <20201014215319.GF2974@worktop.programming.kicks-ass.net>
 <20201014221152.GS3249@paulmck-ThinkPad-P72>
 <20201014223954.GH2594@hirez.programming.kicks-ass.net>
 <20201014235553.GU3249@paulmck-ThinkPad-P72>
 <20201015034128.GA10260@paulmck-ThinkPad-P72>
 <20201015094926.GY2611@hirez.programming.kicks-ass.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20201015094926.GY2611@hirez.programming.kicks-ass.net>
Precedence: bulk

On Thu, Oct 15, 2020 at 11:49:26AM +0200, Peter Zijlstra wrote:
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -1143,13 +1143,15 @@ bool rcu_lockdep_current_cpu_online(void
>  	struct rcu_data *rdp;
>  	struct rcu_node *rnp;
>  	bool ret = false;
> +	unsigned long seq;
>  
>  	if (in_nmi() || !rcu_scheduler_fully_active)
>  		return true;
>  	preempt_disable_notrace();
>  	rdp = this_cpu_ptr(&rcu_data);
>  	rnp = rdp->mynode;
> -	if (rdp->grpmask & rcu_rnp_online_cpus(rnp))
> +	seq = READ_ONCE(rnp->ofl_seq) & ~0x1;
> +	if (rdp->grpmask & rcu_rnp_online_cpus(rnp) || seq != READ_ONCE(rnp->ofl_seq))
>  		ret = true;
>  	preempt_enable_notrace();
>  	return ret;

Also, here, are the two loads important? Wouldn't:

	|| READ_ONCE(rnp->ofl_seq) & 0x1

be sufficient?