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:   Fri, 9 Aug 2019 23:52:22 -0400
From:   Joel Fernandes <joel@joelfernandes.org>
To:     "Paul E. McKenney" <paulmck@linux.ibm.com>
Cc:     Byungchul Park <byungchul.park@lge.com>,
        linux-kernel@vger.kernel.org, Rao Shoaib <rao.shoaib@oracle.com>,
        max.byungchul.park@gmail.com, kernel-team@android.com,
        kernel-team@lge.com, Davidlohr Bueso <dave@stgolabs.net>,
        Josh Triplett <josh@joshtriplett.org>,
        Lai Jiangshan <jiangshanlai@gmail.com>,
        Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
        rcu@vger.kernel.org, Steven Rostedt <rostedt@goodmis.org>
Subject: Re: [PATCH RFC v1 1/2] rcu/tree: Add basic support for kfree_rcu
 batching
Message-ID: <20190810035222.GA157218@google.com>
References: <20190808095232.GA30401@X58A-UD3R>
 <20190808125607.GB261256@google.com>
 <20190808233014.GA184373@google.com>
 <20190809151619.GD28441@linux.ibm.com>
 <20190809153924.GB211412@google.com>
 <20190809163346.GF28441@linux.ibm.com>
 <20190809202226.GC255533@google.com>
 <20190809204217.GN28441@linux.ibm.com>
 <20190809213643.GG255533@google.com>
 <20190810034027.GR28441@linux.ibm.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20190810034027.GR28441@linux.ibm.com>
User-Agent: Mutt/1.10.1 (2018-07-13)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Aug 09, 2019 at 08:40:27PM -0700, Paul E. McKenney wrote:
[snip]
> > > In contrast, a heavy duty userspace-driven workload would transition to
> > > and from userspace for each kfree_rcu(), and that would increment the
> > > dyntick-idle count on each transition to and from userspace.  Adding the
> > > rcu_momentary_dyntick_idle() emulates a pair of such transitions.
> > 
> > But even if we're in kernel mode and not transitioning, I thought the FQS
> > loop (rcu_implicit_dynticks_qs() function) would set need_heavy_qs to true at
> > 2 * jiffies_to_sched_qs.
> > 
> > Hmm, I forgot that jiffies_to_sched_qs can be quite large I guess. You're
> > right, we could call rcu_momentary_dyntick_idle() in advance before waiting
> > for FQS loop to do the setting of need_heavy_qs.
> > 
> > Or, am I missing something with the rcu_momentary_dyntick_idle() point you
> > made?
> 
> The trick is that rcu_momentary_dyntick_idle() directly increments the
> CPU's dyntick counter, so that the next FQS loop will note that the CPU
> passed through a quiescent state.  No need for need_heavy_qs in this case.

Yes, that's what I also understand. Thanks for confirming,

 - Joel