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;
Date:   Tue, 26 Jun 2018 15:27:47 -0700
From:   "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     linux-kernel@vger.kernel.org, mingo@kernel.org,
        jiangshanlai@gmail.com, dipankar@in.ibm.com,
        akpm@linux-foundation.org, mathieu.desnoyers@efficios.com,
        josh@joshtriplett.org, tglx@linutronix.de, rostedt@goodmis.org,
        dhowells@redhat.com, edumazet@google.com, fweisbec@gmail.com,
        oleg@redhat.com, joel@joelfernandes.org
Subject: Re: [PATCH tip/core/rcu 16/27] rcu: Add comment documenting how
 rcu_seq_snap works
Reply-To: paulmck@linux.vnet.ibm.com
References: <20180626003448.GA26209@linux.vnet.ibm.com>
 <20180626003513.27812-16-paulmck@linux.vnet.ibm.com>
 <20180626171454.GI2494@hirez.programming.kicks-ass.net>
 <20180626180855.GE3593@linux.vnet.ibm.com>
 <20180626192113.GM2494@hirez.programming.kicks-ass.net>
 <20180626193146.GK3593@linux.vnet.ibm.com>
 <20180626201503.GS2494@hirez.programming.kicks-ass.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20180626201503.GS2494@hirez.programming.kicks-ass.net>
User-Agent: Mutt/1.5.21 (2010-09-15)
Message-Id: <20180626222747.GP3593@linux.vnet.ibm.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, Jun 26, 2018 at 10:15:03PM +0200, Peter Zijlstra wrote:
> On Tue, Jun 26, 2018 at 12:31:46PM -0700, Paul E. McKenney wrote:
> > > Then the existing comment is misleading and really needs change.
> > 
> > Would it be sufficient to add something like "The memory barrier is
> > required to support the many-to-many ordering guaranteed by RCU grace
> > periods"?
> 
> What would lead me to wonder why after the load and not (also) before.

The memory barriers on the other sides of these primitives are supplied
by the caller.  For example, in srcu_gp_end(), there is a set of
spin_unlock_irq_rcu_node()/spin_lock_irq_rcu_node() pairs between the
rcu_seq_end() of the prior SRCU grace period and the rcu_seq_start()
of the next one.

However, these things aren't anywhere near a fastpath, so I could add
the smp_mb() calls on the other sides for readability, if that would
be useful.

							Thanx, Paul